U.S. patent application number 14/833421 was filed with the patent office on 2017-03-02 for virtual machine placement in a cloud computing environment based on factors including optimized processor-memory affinity.
The applicant listed for this patent is International Business Machines Corporation. Invention is credited to Daniel C. Birkestrand, Peter J. Heyrman, Edward C. Prosser.
Application Number | 20170060627 14/833421 |
Document ID | / |
Family ID | 58095497 |
Filed Date | 2017-03-02 |
United States Patent
Application |
20170060627 |
Kind Code |
A1 |
Birkestrand; Daniel C. ; et
al. |
March 2, 2017 |
VIRTUAL MACHINE PLACEMENT IN A CLOUD COMPUTING ENVIRONMENT BASED ON
FACTORS INCLUDING OPTIMIZED PROCESSOR-MEMORY AFFINITY
Abstract
Optimized placement of virtual machines in a cloud environment
is based on factors that include processor-memory affinity. A smart
migration mechanism (SMM) predicts an optimization score for
multiple permutations of placing virtual machines on a target
system to create an optimal move list. The optimization score is a
theoretical score calculated using dynamic platform optimization
(DPO). The SMM may allow the user to set initial parameters and
change the parameters to create potential changes lists. The move
lists are ranked to allow the user to select the optimal change
list to provide the best affinity, quickest fulfillment of
requirements and least disruption for a given set of
parameters.
Inventors: |
Birkestrand; Daniel C.;
(Rochester, MN) ; Heyrman; Peter J.; (Rochester,
MN) ; Prosser; Edward C.; (Rochester, MN) |
|
Applicant: |
Name |
City |
State |
Country |
Type |
International Business Machines Corporation |
Armonk |
NY |
US |
|
|
Family ID: |
58095497 |
Appl. No.: |
14/833421 |
Filed: |
August 24, 2015 |
Current U.S.
Class: |
1/1 |
Current CPC
Class: |
G06F 9/5088 20130101;
G06F 9/5044 20130101; G06F 2009/4557 20130101; G06F 9/4856
20130101; G06F 9/5016 20130101; G06F 2009/45595 20130101; G06F
9/45558 20130101; G06F 9/5033 20130101 |
International
Class: |
G06F 9/48 20060101
G06F009/48; G06F 9/455 20060101 G06F009/455; H04L 29/08 20060101
H04L029/08 |
Claims
1-9. (canceled)
10. A computer-implemented method executed by at least one
processor for migrating a virtual machine from a source system to a
target system, the method comprising: determining initial
parameters for a plurality of virtual machines on a plurality of
source and target systems; gathering affinity scores and
configuration data from the sources systems and from the target
systems; and generating multiple move lists wherein each move list
comprises one permutation of an ordered list for placing the
plurality of virtual machines on the target system; predicting an
optimization score for each of the move lists to determine an
optimal move list having the best optimization score; and using the
optimal move list to move the at least one virtual machine of the
plurality of virtual machines from the source system to the target
system.
11. The method of claim 10 further comprising predicting the
optimization score using dynamic platform optimization.
12. The method of claim 10 wherein the optimization score is
determined with primary and secondary factors, wherein the primary
factors include processor memory affinity of the virtual machines
placed on the target systems.
13. The method of claim 10 further comprising creating a no-move
list that is referenced when scoring the multiple move lists to
create the optimal move list.
14. The method of claim 10 further comprising giving a user a
plurality of ranked move lists with scores for the ranked move
lists and allows the user to select a move list from the plurality
of ranked move lists to use as the optimal move list.
15. The method of claim 10 further comprising ranking the multiple
move lists and providing a user with a plurality of ranked move
lists with scores for each of the ranked move lists, and allowing
the user to change initial parameters, and when the user changes
parameters, repeating the step of predicting an optimization score
for the multiple move lists.
16. The method of claim 10 further comprising determining no move
lists are needed and performing simple virtual machine
migrations.
17. The method of claim 10 further comprising migrating virtual
machines to target systems according to the optimal move list.
18. The method of claim 10 further comprising migrating a virtual
machine to a system that causes a lower priority virtual machine to
drop below a minimum score but the migration also causes a higher
priority virtual machine a significant increase in score that
results in an overall increase in system performance.
19. A computer-implemented method executed by at least one
processor for migrating a virtual machine from a source system to a
target system, the method comprising: determining initial
parameters for a plurality of virtual machines on a plurality of
source and target systems; gathering affinity scores and
configuration data from the sources systems and from the target
systems; and generating multiple move lists wherein each move list
comprises one permutation of an ordered list for placing the
plurality of virtual machines on the target systems; predicting an
optimization score for each of the move lists using dynamic
platform optimization to determine a plurality of ranked move lists
with scores for the ranked move lists, wherein the optimization
score is determined with primary and secondary factors, wherein the
primary factors include processor memory affinity of the virtual
machines placed on the target systems; giving a user the plurality
of ranked move lists and allowing the user to select a move list
from the plurality of ranked move lists to use as the optimal move
list; and migrating the at least one virtual machine of the
plurality of virtual machines from the source system to the target
system according to the optimal move list selected by the user.
20. The method of claim 19 further comprising allowing the user to
change initial parameters, and when the user changes initial
parameters, repeating the step of predicting an optimization score
for the multiple move lists.
Description
BACKGROUND
[0001] 1. Technical Field
[0002] This invention generally relates to cloud computing systems,
and more specifically relates to placing virtual machines in a
cloud environment based on factors that include optimized
processor-memory affinity.
[0003] 2. Background Art
[0004] Cloud computing is a common expression for distributed
computing over a network and can also be used with reference to
network-based services such as Infrastructure as a Service (IaaS).
IaaS is a cloud based service that provides physical processing
resources to run virtual machines (VMs) as a guest for different
customers. The virtual machine may host a user application or a
server.
[0005] It is often necessary or desirable to migrate workload in
one computer system (a source) to another computer system (a
target). Often, workload migration takes the form of migrating one
or more virtual machines (sometimes referred to as logical
partitions) from the source to the target, the migrated virtual
machine's workload previously being performed in the source being
subsequently performed in the target. For example, each client of a
server may have its own virtual machine within the server for one
or more respective client processes, so the workload is migrated by
moving the workload of one or more clients, and reconstructing the
virtual machine parameters, on one or more other server systems. A
virtual machine may be migrated to balance workload among multiple
systems, but may also be migrated to perform maintenance on the
source system or for some other reason.
[0006] Physically, many large server systems are designed as
systems having a non-uniform memory access in which multiple
processors and main memory are physically distributed, so that each
processor has some portion of main memory which is in closer
physical proximity (and is accessed faster) than other portions of
main memory. In such a system, it is desirable, insofar as
possible, to hold instructions and other data required for
executing a process or thread in the main memory portion which is
physically closest to the processor executing the process or
thread, a characteristic referred to as "processor-memory affinity"
or "affinity".
BRIEF SUMMARY
[0007] An apparatus and method places virtual machines in a cloud
environment based on factors where a primary factor is optimized
processor-memory affinity. A smart migration mechanism (SMM)
predicts an optimization score for multiple permutations of placing
virtual machines on a target system to create an optimal move list.
The optimization score is a theoretical score calculated using
dynamic platform optimization (DPO). The SMM may allow the user to
set initial parameters and change the parameters to create
potential changes lists. The move lists are ranked to allow the
user to select the optimal change list to provide the best
affinity, quickest fulfillment of requirements and least disruption
for a given set of parameters.
[0008] The foregoing and other features and advantages of the
invention will be apparent from the following more particular
description of preferred embodiments of the invention, as
illustrated in the accompanying drawings.
BRIEF DESCRIPTION OF THE SEVERAL VIEWS OF THE DRAWING(S)
[0009] The disclosure will be described in conjunction with the
appended drawings, where like designations denote like elements,
and:
[0010] FIG. 1 is a block diagram of a cloud computing node;
[0011] FIG. 2 is a block diagram of a cloud computing
environment;
[0012] FIG. 3 is a block diagram of abstraction model layers;
[0013] FIG. 4 is a block diagram that illustrates an example a
computer system that includes hardware and memory that could be
allocated to different virtual machines;
[0014] FIG. 5 illustrates manager code and virtual machines in one
of the server systems 203 shown in FIG. 2;
[0015] FIG. 6 illustrates a block diagram of a system that places
virtual machines in a cloud environment based on optimized
processor-memory affinity;
[0016] FIGS. 7 and 8 illustrate an example of creating an optimal
move list for placing virtual machines in a cloud environment based
on optimized processor-memory affinity;
[0017] FIG. 9 is a flow diagram for placing virtual machines in a
cloud environment based on optimized processor-memory affinity;
[0018] FIG. 10 is a flow diagram of an example method for step 870
in FIG. 8; and
[0019] FIG. 11 is a flow diagram of an example method for step 1020
in FIG. 10.
DETAILED DESCRIPTION
[0020] The claims and disclosure herein describe placing virtual
machines in a cloud environment based on factors where a primary
factor is optimized processor-memory affinity. A smart migration
mechanism (SMM) predicts an optimization score for multiple
permutations of placing virtual machines on a target system to
create an optimal move list. The optimization score is a
theoretical score calculated using dynamic platform optimization
(DPO). The SMM may allow the user to set initial parameters and
change the parameters to create potential changes lists. The move
lists are ranked to allow the user to select the optimal change
list to provide the best affinity, quickest fulfillment of
requirements and least disruption for a given set of
parameters.
[0021] It is understood in advance that although this disclosure
includes a detailed description on cloud computing, implementation
of the teachings recited herein are not limited to a cloud
computing environment. Rather, embodiments of the present invention
are capable of being implemented in conjunction with any other type
of computing environment now known or later developed.
[0022] Cloud computing is a model of service delivery for enabling
convenient, on-demand network access to a shared pool of
configurable computing resources (e.g. networks, network bandwidth,
servers, processing, memory, storage, applications, virtual
machines, and services) that can be rapidly provisioned and
released with minimal management effort or interaction with a
provider of the service. This cloud model may include at least five
characteristics, at least three service models, and at least four
deployment models.
[0023] Characteristics are as follows:
[0024] On-demand self-service: a cloud consumer can unilaterally
provision computing capabilities, such as server time and network
storage, as needed automatically without requiring human
interaction with the service's provider.
[0025] Broad network access: capabilities are available over a
network and accessed through standard mechanisms that promote use
by heterogeneous thin or thick client platforms (e.g., mobile
phones, laptops, and PDAs).
[0026] Resource pooling: the provider's computing resources are
pooled to serve multiple consumers using a multi-tenant model, with
different physical and virtual resources dynamically assigned and
reassigned according to demand. There is a sense of location
independence in that the consumer generally has no control or
knowledge over the exact location of the provided resources but may
be able to specify location at a higher level of abstraction (e.g.,
country, state, or datacenter).
[0027] Rapid elasticity: capabilities can be rapidly and
elastically provisioned, in some cases automatically, to quickly
scale out and rapidly released to quickly scale in. To the
consumer, the capabilities available for provisioning often appear
to be unlimited and can be purchased in any quantity at any
time.
[0028] Measured service: cloud systems automatically control and
optimize resource use by leveraging a metering capability at some
level of abstraction appropriate to the type of service (e.g.,
storage, processing, bandwidth, and active user accounts). Resource
usage can be monitored, controlled, and reported providing
transparency for both the provider and consumer of the utilized
service.
[0029] Service Models are as follows:
[0030] Software as a Service (SaaS): the capability provided to the
consumer is to use the provider's applications running on a cloud
infrastructure. The applications are accessible from various client
devices through a thin client interface such as a web browser
(e.g., web-based email). The consumer does not manage or control
the underlying cloud infrastructure including network, servers,
operating systems, storage, or even individual application
capabilities, with the possible exception of limited user-specific
application configuration settings.
[0031] Platform as a Service (PaaS): the capability provided to the
consumer is to deploy onto the cloud infrastructure
consumer-created or acquired applications created using programming
languages and tools supported by the provider. The consumer does
not manage or control the underlying cloud infrastructure including
networks, servers, operating systems, or storage, but has control
over the deployed applications and possibly application hosting
environment configurations.
[0032] Infrastructure as a Service (IaaS): the capability provided
to the consumer is to provision processing, storage, networks, and
other fundamental computing resources where the consumer is able to
deploy and run arbitrary software, which can include operating
systems and applications. The consumer does not manage or control
the underlying cloud infrastructure but has control over operating
systems, storage, deployed applications, and possibly limited
control of select networking components (e.g., host firewalls).
[0033] Deployment Models are as follows:
[0034] Private cloud: the cloud infrastructure is operated solely
for an organization. It may be managed by the organization or a
third party and may exist on-premises or off-premises.
[0035] Community cloud: the cloud infrastructure is shared by
several organizations and supports a specific community that has
shared concerns (e.g., mission, security requirements, policy, and
compliance considerations). It may be managed by the organizations
or a third party and may exist on-premises or off-premises.
[0036] Public cloud: the cloud infrastructure is made available to
the general public or a large industry group and is owned by an
organization selling cloud services.
[0037] Hybrid cloud: the cloud infrastructure is a composition of
two or more clouds (private, community, or public) that remain
unique entities but are bound together by standardized or
proprietary technology that enables data and application
portability (e.g., cloud bursting for loadbalancing between
clouds).
[0038] A cloud computing environment is service oriented with a
focus on statelessness, low coupling, modularity, and semantic
interoperability. At the heart of cloud computing is an
infrastructure comprising a network of interconnected nodes.
[0039] Referring now to FIG. 1, a block diagram of an example of a
cloud computing node is shown. Cloud computing node 100 is only one
example of a suitable cloud computing node and is not intended to
suggest any limitation as to the scope of use or functionality of
embodiments of the invention described herein. Regardless, cloud
computing node 100 is capable of being implemented and/or
performing any of the functionality set forth hereinabove.
[0040] In cloud computing node 100 there is a computer
system/server 110, which is operational with numerous other general
purpose or special purpose computing system environments or
configurations. Examples of well-known computing systems,
environments, and/or configurations that may be suitable for use
with computer system/server 110 include, but are not limited to,
personal computer systems, server computer systems, thin clients,
thick clients, handheld or laptop devices, multiprocessor systems,
microprocessor-based systems, set top boxes, programmable consumer
electronics, network PCs, minicomputer systems, mainframe computer
systems, and distributed cloud computing environments that include
any of the above systems or devices, and the like.
[0041] Computer system/server 110 may be described in the general
context of computer system executable instructions, such as program
modules, being executed by a computer system. Generally, program
modules may include routines, programs, objects, components, logic,
data structures, and so on that perform particular tasks or
implement particular abstract data types. Computer system/server
110 may be practiced in distributed cloud computing environments
where tasks are performed by remote processing devices that are
linked through a communications network. In a distributed cloud
computing environment, program modules may be located in both local
and remote computer system storage media including memory storage
devices.
[0042] As shown in FIG. 1, computer system/server 110 in cloud
computing node 100 is shown in the form of a general-purpose
computing device. The components of computer system/server 110 may
include, but are not limited to, one or more processors or
processing units 120, a system memory 130, and a bus 122 that
couples various system components including system memory 130 to
processing unit 120.
[0043] Bus 122 represents one or more of any of several types of
bus structures, including a memory bus or memory controller, a
peripheral bus, an accelerated graphics port, and a processor or
local bus using any of a variety of bus architectures. By way of
example, and not limitation, such architectures include Industry
Standard Architecture (ISA) bus, Micro Channel Architecture (MCA)
bus, Enhanced ISA (EISA) bus, Video Electronics Standards
Association (VESA) local bus, and Peripheral Component Interconnect
(PCI) bus.
[0044] Computer system/server 110 typically includes a variety of
computer system readable media. Such media may be any available
media that is accessible by computer system/server 110, and it
includes both volatile and non-volatile media, removable and
non-removable media. Examples of removable media are shown in FIG.
1 to include a Digital Video Disc (DVD) 192.
[0045] System memory 130 can include computer system readable media
in the form of volatile or non-volatile memory, such as firmware
132. Firmware 132 provides an interface to the hardware of computer
system/server 110. System memory 130 can also include computer
system readable media in the form of volatile memory, such as
random access memory (RAM) 134 and/or cache memory 136. Computer
system/server 110 may further include other
removable/non-removable, volatile/non-volatile computer system
storage media. By way of example only, storage system 140 can be
provided for reading from and writing to a non-removable,
non-volatile magnetic media (not shown and typically called a "hard
drive"). Although not shown, a magnetic disk drive for reading from
and writing to a removable, non-volatile magnetic disk (e.g., a
"floppy disk"), and an optical disk drive for reading from or
writing to a removable, non-volatile optical disk such as a CD-ROM,
DVD-ROM or other optical media can be provided. In such instances,
each can be connected to bus 122 by one or more data media
interfaces. As will be further depicted and described below, memory
130 may include at least one program product having a set (e.g., at
least one) of program modules that are configured to carry out the
functions described in more detail below.
[0046] Program/utility 150, having a set (at least one) of program
modules 152, may be stored in memory 130 by way of example, and not
limitation, as well as an operating system, one or more application
programs, other program modules, and program data. Each of the
operating system, one or more application programs, other program
modules, and program data or some combination thereof, may include
an implementation of a networking environment. Program modules 152
generally carry out the functions and/or methodologies of
embodiments of the invention as described herein.
[0047] Computer system/server 110 may also communicate with one or
more external devices 190 such as a keyboard, a pointing device, a
display 180, a disk drive, etc.; one or more devices that enable a
user to interact with computer system/server 110; and/or any
devices (e.g., network card, modem, etc.) that enable computer
system/server 110 to communicate with one or more other computing
devices. One suitable example of an external device 190 is a DVD
drive which can read a DVD 192 as shown in FIG. 1. Such
communication can occur via Input/Output (I/O) interfaces 170.
Still yet, computer system/server 110 can communicate with one or
more networks such as a local area network (LAN), a general wide
area network (WAN), and/or a public network (e.g., the Internet)
via network adapter 160. As depicted, network adapter 160
communicates with the other components of computer system/server
110 via bus 122. It should be understood that although not shown,
other hardware and/or software components could be used in
conjunction with computer system/server 110. Examples, include, but
are not limited to: microcode, device drivers, redundant processing
units, external disk drive arrays, Redundant Array of Independent
Disk (RAID) systems, tape drives, data archival storage systems,
etc.
[0048] Referring now to FIG. 2, illustrative cloud computing
environment 200 is depicted. As shown, cloud computing environment
200 comprises one or more cloud computing nodes 100 with which
local computing devices used by cloud consumers, such as, for
example, personal digital assistant (PDA) or cellular telephone
210A, desktop computer 210B, laptop computer 210C, and/or
automobile computer system 210N may communicate. Nodes 100 may
communicate with one another. They may be grouped (not shown)
physically or virtually, in one or more networks, such as Private,
Community, Public, or Hybrid clouds as described hereinabove, or a
combination thereof. This allows cloud computing environment 200 to
offer infrastructure, platforms and/or software as services for
which a cloud consumer does not need to maintain resources on a
local computing device. It is understood that the types of
computing devices 210A-N shown in FIG. 2 are intended to be
illustrative only and that computing nodes 100 and cloud computing
environment 200 can communicate with any type of computerized
device over any type of network and/or network addressable
connection (e.g., using a web browser).
[0049] Again referring to FIG. 2, the cloud computing environment
preferably includes a hardware management console 202 to manage the
cloud computing nodes 100 including one or more server computer
systems 203 (herein generically referred to as feature 203).
Hardware management console 202 and server computer systems 203 are
preferably general purpose digital computers, each having a
respective at least one programmable central processing unit (CPU)
which executes instructions storable in an addressable memory such
as the computer system/server 110 illustrated in FIG. 1. Digital
devices may further include one or more storage servers 204A-204B
(herein generically referred to as feature 204) which function as
shared data storage available to server computer systems 203. The
networked environment may further include additional devices (not
shown), such as routers and special purpose digital devices for
performing accounting, maintenance, backup, and other
functions.
[0050] Hardware management console 202 supports an interactive user
interface enabling a system administrator or similar user to manage
allocations of resources among the various digital data devices, in
particular servers 203. In particular, in accordance with one or
more embodiments, hardware management console 202 manages the
migration of logical partitions or virtual machines from one server
203 to another, as described more fully herein. Hardware management
console 202 may further perform other functions of managing a
network of servers, such as providing a portal for client requests,
assigning client requests to servers and/or logical partitions
therein for execution, managing maintenance operations, configuring
network connections, and so forth.
[0051] Although illustrated as a stand-alone device attached
directly to network 201, hardware management console 202 may
alternatively be implemented as a software program executing in one
of servers 203 (preferably in its own logical partition) to which
an interactive terminal is directly attached, or which is accessed
by a remote terminal over network 201. The multiple computer system
networked environment may include only a single hardware management
console 202 as shown in FIG. 2, but may alternatively include
multiple hardware management consoles which collectively support
the multiple computer systems and share the tasks of managing
allocation of resources, etc. Multiple hardware management consoles
202 provide redundancy and continuous operation in the event that
any one console malfunctions. In the present Specification,
hardware management console 202 will be described as a single
computer system apart from the server systems 203, for simplicity
and clarity of description, it being understood that the hardware
and software components of a hardware management console 202 and
their operation described herein may be embodied in and performed
by one or more physical server computer systems 203 and software
executing thereon.
[0052] Referring now to FIG. 3, a set of functional abstraction
layers provided by cloud computing environment 200 (FIG. 2) is
shown. It should be understood in advance that the components,
layers, and functions shown in FIG. 3 are intended to be
illustrative only and the disclosure and claims are not limited
thereto. As depicted, the following layers and corresponding
functions are provided.
[0053] Hardware and software layer 310 includes hardware and
software components. Examples of hardware components include
mainframes 352; RISC (Reduced Instruction Set Computer)
architecture based servers 354; servers 356; blade servers 358;
storage devices 360; and networks and networking components 362. In
some embodiments, software components include network application
server software 364 and database software 366.
[0054] Virtualization layer 320 provides an abstraction layer from
which the following examples of virtual entities may be provided:
virtual servers 368; virtual storage 370; virtual networks 372,
including virtual private networks; virtual applications and
operating systems 374; and virtual clients 376.
[0055] In one example, management layer 330 may provide the
functions described below. Resource provisioning 378 provides
dynamic procurement of computing resources and other resources that
are utilized to perform tasks within the cloud computing
environment. Metering and Pricing 380 provide cost tracking as
resources are utilized within the cloud computing environment, and
billing or invoicing for consumption of these resources. In one
example, these resources may comprise application software
licenses. Security provides identity verification for cloud
consumers and tasks, as well as protection for data and other
resources. User portal 382 provides access to the cloud computing
environment for consumers and system administrators. Service level
management 384 provides cloud computing resource allocation and
management such that required service levels are met. Service Level
Agreement (SLA) planning and fulfillment 386 provide
pre-arrangement for, and procurement of, cloud computing resources
for which a future requirement is anticipated in accordance with an
SLA. The management layer further includes a smart migration
mechanism (SMM) 350 as described herein. While the SMM 350 is shown
in FIG. 3 to reside in the management layer 330, the SMM 350
actually may span other levels shown in FIG. 3 as needed.
[0056] Workloads layer 340 provides examples of functionality for
which the cloud computing environment may be utilized. Examples of
workloads and functions which may be provided from this layer
include: mapping and navigation 386; software development and
lifecycle management 390; virtual classroom education delivery 392;
data analytics processing 394; transaction processing 396 and
mobile desktop 398.
[0057] As will be appreciated by one skilled in the art, aspects of
this disclosure may be embodied as a system, method or computer
program product. Accordingly, aspects may take the form of an
entirely hardware embodiment, an entirely software embodiment
(including firmware, resident software, micro-code, etc.) or an
embodiment combining software and hardware aspects that may all
generally be referred to herein as a "circuit," "module" or
"system." Furthermore, aspects of the present invention may take
the form of a computer program product embodied in one or more
computer readable medium(s) having computer readable program code
embodied thereon.
[0058] Any combination of one or more computer readable medium(s)
may be utilized. The computer readable medium may be a computer
readable signal medium or a non-transitory computer readable
storage medium. A computer readable storage medium may be, for
example, but not limited to, an electronic, magnetic, optical,
electromagnetic, infrared, or semiconductor system, apparatus, or
device, or any suitable combination of the foregoing. More specific
examples (a non-exhaustive list) of the computer readable storage
medium would include the following: an electrical connection having
one or more wires, a portable computer diskette, a hard disk, a
random access memory (RAM), a read-only memory (ROM), an erasable
programmable read-only memory (EPROM or Flash memory), an optical
fiber, a portable compact disc read-only memory (CD-ROM), an
optical storage device, a magnetic storage device, or any suitable
combination of the foregoing. In the context of this document, a
computer readable storage medium may be any tangible medium that
can contain, or store a program for use by or in connection with an
instruction execution system, apparatus, or device.
[0059] A computer readable signal medium may include a propagated
data signal with computer readable program code embodied therein,
for example, in baseband or as part of a carrier wave. Such a
propagated signal may take any of a variety of forms, including,
but not limited to, electro-magnetic, optical, or any suitable
combination thereof. A computer readable signal medium may be any
computer readable medium that is not a computer readable storage
medium and that can communicate, propagate, or transport a program
for use by or in connection with an instruction execution system,
apparatus, or device.
[0060] Program code embodied on a computer readable medium may be
transmitted using any appropriate medium, including but not limited
to wireless, wireline, optical fiber cable, RF, etc., or any
suitable combination of the foregoing.
[0061] Computer program code for carrying out operations for
aspects of the present invention may be written in any combination
of one or more programming languages, including an object oriented
programming language such as Java, Smalltalk, C++ or the like and
conventional procedural programming languages, such as the "C"
programming language or similar programming languages. The program
code may execute entirely on the user's computer, partly on the
user's computer, as a stand-alone software package, partly on the
user's computer and partly on a remote computer or entirely on the
remote computer or server. In the latter scenario, the remote
computer may be connected to the user's computer through any type
of network, including a local area network (LAN) or a wide area
network (WAN), or the connection may be made to an external
computer (for example, through the Internet using an Internet
Service Provider).
[0062] Aspects of the present invention are described below with
reference to flowchart illustrations and/or block diagrams of
methods, apparatus (systems) and computer program products
according to embodiments of the invention. It will be understood
that each block of the flowchart illustrations and/or block
diagrams, and combinations of blocks in the flowchart illustrations
and/or block diagrams, can be implemented by computer program
instructions. These computer program instructions may be provided
to a processor of a general purpose computer, special purpose
computer, or other programmable data processing apparatus to
produce a machine, such that the instructions, which execute via
the processor of the computer or other programmable data processing
apparatus, create means for implementing the functions/acts
specified in the flowchart and/or block diagram block or
blocks.
[0063] These computer program instructions may also be stored in a
computer readable medium that can direct a computer, other
programmable data processing apparatus, or other devices to
function in a particular manner, such that the instructions stored
in the computer readable medium produce an article of manufacture
including instructions which implement the function/act specified
in the flowchart and/or block diagram block or blocks.
[0064] The computer program instructions may also be loaded onto a
computer, other programmable data processing apparatus, or other
devices to cause a series of operational steps to be performed on
the computer, other programmable apparatus or other devices to
produce a computer implemented process such that the instructions
which execute on the computer or other programmable apparatus
provide processes for implementing the functions/acts specified in
the flowchart and/or block diagram block or blocks.
[0065] As will be appreciated by one skilled in the art, aspects of
the present invention may be embodied as a system, method or
computer program product. Accordingly, aspects of the present
invention may take the form of an entirely hardware embodiment, an
entirely software embodiment (including firmware, resident
software, micro-code, etc.) or an embodiment combining software and
hardware aspects that may all generally be referred to herein as a
"circuit," "module" or "system." Furthermore, aspects of the
present invention may take the form of a computer program product
embodied in one or more computer readable medium(s) having computer
readable program code embodied thereon.
DEFINITIONS
[0066] Processor-memory Affinity--generally describes how
physically close allocated processors are to associated physical
memory to improve performance.
[0067] Affinity Scores--a rating for processor-memory affinity for
a virtual machine, a system or group of systems.
[0068] CiP--Contain in Primary is used with affinity evaluation and
scoring where processors and memory fit within a chip. A primary
domain is a collection of processors and associated memory locally
contained on a physical chip. A primary domain on a chip is the
densest physical attribute that affects affinity.
[0069] CiS--Contain in Secondary is used with affinity evaluation
and scoring where processors and memory cannot fit within a chip,
but must be spread across multiple chips, but can fit within a
secondary domain.
[0070] DPO--Dynamic Platform Optimization attempts to improve
virtual machine performance by re-assigning processors and memory
among running or powered off virtual machines to improve
processor-memory affinity for some or all of them. DPO sometimes
refers to a `DPO operation` if actually re-assigning resources. DPO
can also refer to a DPO score. A DPO score is generated by
implementing much of the up front DPO function for planned resource
moves, but not actually following through with re-assigning
anything (the scoring algorithm needs to be aware of what moves are
possible just as the DPO moves functionality does).
[0071] DPO Score--A DPO valule (for example 0 to 100, 0 worst, 100
best) that is determined using algorithms that use a set of rules
(CiP, CiS, etc.) governing the allowed physical arrangement of
processors and memory on a server or group of servers, the virtual
machine requirements for current and max resources, and the
possible allocation mappings that can occur given the physical
limitations in a server or group of servers. DPO scoring can be
done against a current configuration, but also a theoretical
potential configuration (to compare before and after effects). A
low current score, and a higher theoretical potential score, would
be the primary reason to attempt a DPO operation to re-assign
resources.
[0072] HMC--Hardware Management Console, a HW/SW device that
contains a GUI and command line interface used for management of a
server or group of servers. The HMC allows a user to perform system
or virtual machine power on/off, resource allocations, virtual
machine definition, virtual machine migration, DPO, licensing, etc.
The SMM function 350 described above may be part of the HMC.
[0073] Hypervisor--A low level software layer that executes
directly on system processors and manages the dispatching of
processors and allocation of memory for the virtual machines
(prevents virtual machines from using each others resources
inadvertently), among other functions such as I/O management,
mobility, and asset protection.
[0074] LPM--Live Partition (virtual machine) Mobility,
functionality on both source and target systems that offloads a
running virtual machine to another host system, with essentially no
disruption to the users for the purpose of consolidation,
maintenance, or performance improvement etc.
[0075] Mobile CoD--Mobile Capacity on Demand, it's a version of
licensing that allows the temporary expansion of processor and/or
memory resources to a pool of servers, whereas traditional
varieties of CoD deal with a single server. This allows for easier
movement of virtual machines within a shared pool using LPM in a
private cloud environment--customers do not need to get new
hardware licenses, and remove licenses etc anymore if using Mobile
CoD.
[0076] Potential DPO Score--Sometimes called theoretical score,
it's calculated on the same numeric scale as a DPO score, but it
uses a theoretical rearrangement of processors and memory as input
to the scoring algorithm, versus the actual current allocations.
This is used to compare with a current score to determine if it's
worth the costs of doing a migration and/or DPO to improve
performance for a particular virtual machine or system (DPO can
have adverse effects too, since it's purpose is to improve
performance for high priority virtual machines, it in turn can be
detrimental to lower priority virtual machines in addition to the
system performance impact during the DPO itself).
[0077] SaC--Spread across Cluster, term used with affinity
evaluation and scoring if processors and memory cannot fit within a
single system, so the virtual machine uses resources on more than
one physical system in a pool or cloud at the same time. Individual
systems involved in a SaC spread would be using any of the other
spread types (such as CiP, etc) within that system.
[0078] SaS--Spread across Secondary, as used with affinity
evaluation and scoring, where processors and memory cannot fit
within one domain, they must be spread across multiple domains.
[0079] WiF--Wherever it Fits, in affinity evaluation and scoring
where processors and memory cannot be spread evenly across domains
(perhaps some drawers don't have memory, etc), then just place it
where it will fit and spread as evenly as posssible across
chips.
[0080] Referring to FIG. 4, a computer system 400 is shown as an
example of a computer system that includes hardware and memory that
could be allocated to different virtual machines. The hardware is
divided into a hardware domain hierarchy according to the physical
boundaries of the hardware. We assume for this specific example
that primary domains correspond to chips, while secondary domains
correspond to nodes or drawers. Note, however, the concepts of
primary domains and secondary domains could be applied to any
suitable hardware hierarchy, whether currently known or developed
in the future. Computer system 400 includes four secondary domains
402A, 402B, 402C and 402D. Each secondary domain includes two
primary domains. Thus, secondary domain 402A includes primary
domains 404A and 404B; secondary domain 402B includes primary
domains 404C and 404D; secondary domain 402C includes primary
domains 404E and 404F; and secondary domain 402D includes primary
domains 404G and 404H. Each primary domain has physical processors
and physical memory. Thus, primary domain 404A includes four
processors 410A and 16 gigabytes (GB) of memory 420A; primary
domain 404B includes four processors 410B and 16 GB of memory 420B;
primary domain 404C includes two processors 410C and 16 GB of
memory 420C; primary domain 404D includes two processors 410D and
48 GB of memory 420D; primary domain 404E includes eight processors
410E and 32 GB of memory 420E; primary domain 404F includes eight
processors 410F and 32 GB of memory 420F; primary domain 404G
includes two processors 410G and 32 GB of memory 420G; and primary
domain 404H includes two processors 410H and 64 GB of memory
420H.
[0081] Logical Partitioning of Computer System Resources
[0082] In the preferred embodiment, each server system 203 is
logically partitionable into a plurality of virtual machines each
executing on behalf of a respective client or performing
administrative or other functions. Partitioning is a technique for
dividing a single large computer system into multiple partitions or
virtual machines, each of which behaves in some respects as a
separate computer system. Computer system resources may be
allocated in any of various ways for use by the virtual machines. A
given resource may be allocated for exclusive use by a single
particular virtual machine, or may be shared among all virtual
machines (or some subgroup of virtual machines) on a time
interleaved or other basis. Some resources may be allocated to
respective particular virtual machines, while others are shared.
Examples of resources which may be partitioned are processors (or
CPUs), memory, data storage within storage units, and network
bandwidth. I/O adapters are typically shared, although they could
be partitioned as well. Each client accessing any of servers 203
executes its own tasks in the virtual machine partition assigned to
the client, meaning that it can use only the system resources or
share of resources assigned to that virtual machine, and not
resources assigned to other virtual machines. Additionally, some
virtual machines may be used for administrative, maintenance, and
other functions, in particular the functions of a hardware
management console 202 as described herein.
[0083] Virtual machine partitioning of resources is virtual rather
than physical. Server computer systems 203 preferably have physical
data connections such as buses running among different hardware
components, allowing them to communicate with one another. These
hardware resources may be shared by and/or allocated to different
virtual machines. From a physical configuration standpoint, there
is preferably no distinction made with regard to virtual machine
partitions. The system's physical devices and subcomponents thereof
are preferably physically connected to allow communication without
regard to virtual machine partitions, and from this hardware
standpoint, there is nothing which prevents a task executing in
virtual machine A from writing to memory or storage allocated to
virtual machine B.
[0084] Generally, allocation of resources to a virtual machine is
enforced by a partition manager embodied as low-level encoded
executable instructions and data, although there may be a certain
amount of hardware support for virtual machine partitioning, such
as special hardware registers which hold state information. The
partition manager (and associated hardware, if any) prevent access
by a virtual machine to the resources allocated to another virtual
machine. Code enforcement of partitioning constraints generally
means that it is possible to alter the virtual configuration of a
partitioned computer system, i.e., to change the number of virtual
machines or re-assign resources to different virtual machines,
without reconfiguring hardware. In the preferred embodiment
described herein, this low-level logical partitioning code is
referred to as the "hypervisor".
[0085] FIG. 5 is a conceptual illustration showing the existence of
manager code and virtual machine partitions at different hardware
and software levels of abstraction in one of server systems 203.
FIG. 5 represents a system having four client virtual machines in
respective partitions 504-507, each executing one or more
applications on behalf of the respective client. These are
designated "VM1", "VM2", etc., it being understood that the number
of partitions may vary. As is well known, a computer system is a
sequential state machine which performs processes. These processes
can be represented at varying levels of abstraction. At a high
level of abstraction, a user specifies a process and input, and
receives an output. As one progresses to lower levels, one finds
that these processes are sequences of instructions in some
programming language, which continuing lower are translated into
lower level instruction sequences, and pass through licensed
internal code and ultimately to data bits which get put in machine
registers to force certain actions. At a very low level, changing
electrical potentials cause various transistors to turn on and off.
In FIG. 5, the "higher" levels of abstraction are generally
represented toward the top of the figure, while lower levels are
represented toward the bottom.
[0086] As shown in FIG. 5 and explained earlier, logical
partitioning of machine resources is a code-enforced concept. In
general, at the hardware level 501, partition boundaries do not
exist (although there may be certain special purpose registers or
other hardware used to identify partition boundaries or other
virtual machine aspects). As used herein, hardware level 501
represents the collection of physical devices (as opposed to data
stored in devices), such as processors, memory, buses, I/O devices,
etc., shown in FIG. 3 and FIG. 4, possibly including other hardware
not shown in FIG. 3 or FIG. 4. As far as a processor 402 is
concerned, it is merely executing machine level instructions. While
code can direct tasks in certain virtual machines to execute on
certain processors, there is nothing in the processor itself which
dictates this assignment, and in fact the assignment can be changed
by the code. Therefore the hardware level is represented in FIG. 5
as a single entity 501, which does not itself distinguish among
virtual machines.
[0087] Partition boundaries of the virtual machines are enforced by
a partition manager (also known as a "hypervisor"), consisting of a
non-relocatable, non-dispatchable portion 502, and a relocatable,
dispatchable portion 503. The hypervisor is super-privileged
executable code which is capable of accessing resources, such as
processor resources and memory, assigned to any virtual machine.
The hypervisor maintains state data in various special purpose
hardware registers, and in tables or other structures in general
memory, which govern boundaries and behavior of the virtual
machines. Among other things, this state data defines the
allocation of resources to virtual machines, and the allocation is
altered by changing the state data rather than by physical
reconfiguration of hardware.
[0088] In the preferred embodiment, the non-dispatchable hypervisor
502 comprises non-relocatable instructions which are executed by
any of processors 402 just as instructions for tasks executing in
the virtual machines. The code is non-relocatable, meaning that the
code which constitutes the non-dispatchable hypervisor is at fixed
real addresses in memory. Non-dispatchable hypervisor 502 has
access to the entire real memory address range of the computer
system, and can manipulate real memory addresses. The dispatchable
hypervisor code 503 (as well as all code executing within a virtual
machine) is contained at addresses which are relative to an address
range assigned to the virtual machine in which it executes, and
therefore this code is relocatable. The dispatchable hypervisor
behaves in much the same manner as a client's virtual machine, but
it is hidden from the clients and not available to execute user
applications. In general, non-dispatchable hypervisor 502 handles
assignment of tasks to physical processors, memory mapping and
virtual machine enforcement, and similar essential tasks required
to execute application code in a partitioned system, while
dispatchable hypervisor 503 handles maintenance-oriented tasks,
such as creating and altering virtual machine definitions.
[0089] As represented in FIG. 5, there is no direct path between
higher levels (levels above non-dispatchable hypervisor 502) and
hardware level 501. While machine instructions of tasks executing
at higher levels can execute directly on a processor 402, access to
hardware resources is controlled by the non-dispatchable
hypervisor. Non-dispatchable hypervisor 502 enforces virtual
machine boundaries of processor resources. Task dispatchers at a
higher level (the respective operating systems) dispatch tasks to
virtual processors defined by the virtual machine parameters, and
the hypervisor in turn dispatches virtual processors to physical
processors at the hardware level 501 for execution of the
underlying task. The hypervisor also enforces partitioning of other
resources, such as allocations of memory to partitions, and routing
I/O to I/O devices associated with the proper partition.
[0090] Dispatchable hypervisor 503 performs many auxiliary system
management functions which are not the province of any client
virtual machine. The dispatchable hypervisor generally performs
higher level virtual machine management operations such as creating
and deleting virtual machines, concurrent hardware maintenance,
allocating processors, memory and other hardware resources to
various virtual machines, etc. In particular, in one or more
embodiments dispatchable hypervisor 503 includes a dynamic platform
optimizer utility 508 which dynamically analyzes and adjusts system
configuration parameters, and a migration agent 509 which handles
migration of partitions from one server system 203 to another
responsive to commands from the hardware management console, as
explained in further detail herein.
[0091] Above non-dispatchable hypervisor 502 are a plurality of
virtual machines 504-507. Each virtual machine behaves, from the
perspective of processes executing within it, as an independent
computer system, having its own memory space and other resources,
and for this reason is also referred to as a virtual machine. Each
virtual machine therefore contains a respective operating system
kernel herein identified as the "OS kernel" 511-514. At the level
of the OS kernel and above, each virtual machine behaves
differently, and therefore FIG. 5 represents the OS Kernel as four
different entities 511-514 corresponding to the four different
virtual machines. In general, each OS kernel 511-514 performs
roughly equivalent functions. However, it is not necessarily true
that all OS kernels 511-514 are identical copies of one another,
and they could be different versions of architecturally equivalent
operating systems, or could even be architecturally different
operating systems. OS kernels 511-514 perform a variety of task
management functions, such as task dispatching, paging, enforcing
data integrity and security among multiple tasks, and so forth.
[0092] Above the OS kernels in each respective virtual machine
there may be any of various applications and data 521-524. In
particular, for server systems 203 supporting virtual machines
executing processes on behalf of remote clients 103, these are
applications executing on behalf of the respective clients and
associated data generated or used by those applications.
Additionally, these applications could represent a hardware
management console and associated applications and data, as further
described herein with respect to FIG. 6. Although applications and
data 521-524 have the same appearance in the conceptual
representation of FIG. 5, it will be understood that in general
each virtual machine includes different applications and data.
[0093] Referring to FIG. 6, a block diagram illustrates a system
600 for placing virtual machines in a cloud environment based on
factors such as optimized processor-memory affinity. The smart
migration mechanism (SMM) 350 was introduced above with reference
to FIG. 3. In this example, the SMM 350 resides in memory 602 which
is located in the hardware management console 202 (FIG. 2). The SMM
performs move list calculations to allow a user to select an
optimal move list 610. To calculate the optimal move list, the SMM
first gathers initial parameters 612 from the user. The SMM then
gathers data from a source system 614 with source system
characteristics 616 and from a target server system(s) 618 with
target system(s) characteristics 620. The SMM 350 performs DPO
calculations for multiple permutations of moving virtual machines
to the target server system(s) 618 to determine the optimal move
list 610 as described further below.
[0094] Before the SMM 350 creates the move lists to determine the
optimal move list 610, the SMM first determines the initial
parameters for the list generation. The intial parameters may
include the following parameters provided by the user: [0095] 1. A
list of source and potential target systems in the pool. [0096] 2.
Logical virtual machine priorities (which are more important, or
the same, such as production vs development, or based on function).
[0097] 3. A minimal processor-memory affinity score, or related
performance characteristic, for each virtual machine where final
scores should not transition below this score. [0098] 4. A goal
processor-memory affinity score for each virtual machine that the
SMM will try to achieve. [0099] 5. Security concerns. For example a
production virtual machine may not be allowed to migrate to a
development system etc. For each target system there may also be
other virtual machines defined that aren't designated by the user
(or some automation mechanism) for improvement, so they should also
have minimal affinity scores and similar security information that
will help prevent them from being negatively affected beyond these
defined thresholds. [0100] 6. An overall migration time frame limit
or date/times to finish may be specified. This defines the maximum
time the SMM is allowed to perform the optimization (there may be
windows where moves should be discouraged, such as during peak
usage times or during maintenance windows when systems are
down).
[0101] After getting the initial parameters from the user, the SMM
350 communicates with the source and target systems to gather
needed information to perform initial checks and create the move
lists. This information may include: [0102] 1. virtual machine
configurations including virtual machine processor counts, virtual
machine memory sizes, allocated processor and memory topography
(physical processor and memory relationships to each other) for
each virtual machine. [0103] 2. total processor counts, currently
allocated and available, for each system. [0104] 3. total memory
size, currently allocated and available, for each system. [0105] 4.
processor and memory hardware topography for each system (a system
view vs logical partition view). [0106] 5. current system and
virtual machine affinity scores, noting where virtual machines rank
(against the minimum and goal scores previously gathered.
[0107] After gathering the above data, the SMM 350 has enough
information to perform initial checks and determine if a move list
is needed. The SMM may do initial quick checks to see if all goal
requirements are already met, or if not, by taking the least
disruptive and/or quickest options first that don't involve
migrations. For example, if all the criteria are already met, such
as when virtual machine affinity scores meet or exceed the goal
affinity scores, then the SMM does not need to take further action.
Further, the SMM could determine that all criteria could be met by
just performing DPO on source systems only and no migrations are
needed. Further, if only a few virtual machines need to migrate to
achieve the affinity score goals, then the SMM can use simple
ordering of virtual machine migrations to target systems and DPO is
not needed. In this case, no move list generation is required other
than perhaps migrating the highest priority virtual machine first
and the rest in order of priority. Typically migration is slower
and more disruptive to the virtual machines than DPO. In these
simplified cases, all criteria can be met by performing DPO on
source and/or target systems, and simple virtual machine migrations
to target systems. If all criteria cannot be met using simple
migrations and DPO on the source and target systems, then the SMM
proceeds to use DPO-score predictions for multiple permutations to
determine an optimal move list as described below and then migrates
the virtual machines to the target systems per the optimal move
list.
[0108] Where the SMM 350 determines requirements can not be met
using simple migrations and optimization of the source system, the
SMM proceeds to create the optimal move list 610 using DPO-score
predictions for multiple permutations. To create the optimal move
list, the SMM creates multiple move permutations and performs
DPO-score predictions of placing virtual machines on the target
systems for each of these permutations as described further below.
The SMM may then give the user the various lists of the
permutations ranked by score. There may be more than one ranked
list. The ranking may be by best overall performance, best
factoring in the priority, fastest to completion, least disruptive
to target systems, etc. The user may be given the oportunity to
change initial parameters and the SMM 350 then recreates the ranked
lists with the new initial parameters. Where the user is satisfied
and does not change the parameters, the user is allowed to approve
a move list as the optimal move list for migration of the virtual
machines to the target systems.
[0109] To score the move list permutations to determine the optimal
move list, the SMM performs DPO scoring to predict a score for each
of the permutations. For the DPO-score predictions the SMM may
assign processor-memory affinity scores as known in the prior art.
Typically, the affinity scores used a 0 (worst) to 100 (perfect)
scale for logical virtual machines and for systems. The
processor-memory affinity score is affected by the processor memory
spread, meaning the physical spread between the processor and
memory. The types of processor-memory spreads in general go from
best case to worst case. In general, it is best to have the
processors and memory in use by a virtual machine as physically
close as possible. These spread types take into account common
hardware boundaries that are typically present in today's computer
servers: [0110] 1. Contain in primary (CiP)<---Best [0111] 2.
Contain in secondary (CiS) [0112] 3. Spread across secondary (SaS)
[0113] 4. Wherever it fits (WiF) [0114] 5. Spread across cluster
(SaC)<---Worst
[0115] The SMM may consider multiple factors to perform DPO-score
predictions for multiple permutations to create the optimal move
list 610. Some factors can be given a greater weight than other
factors. The various example factors are described herein for
discussion into primary factors and secondary factors. In the
examples herein, the SMM gives the most weight to the primary
factors and lessor weight to the secondary factors.
[0116] The primary factors for move list generation may include:
[0117] 1. DPO score improvements (The difference between current
and predicted scores). [0118] 2. DPO score degradation (within
minimums).
[0119] The secondary factors for move list generation may include:
[0120] 1. DPO operations on processors (throughput degradation
during re-assignments). [0121] 2. DPO operations on memory
(throughput degradation during re-assignments). [0122] 3. Migration
costs (users experience some slowdown in throughput during live
migration). [0123] 4. Processor and memory licensing (increased or
decreased expense). [0124] 5. Processor and memory types. If the
processors and memory to be assigned to the migrating virtual
machine differ in any significant way that would affect a
cost-benefit analysis (speed, efficiency, mirrored vs non-mirrored,
etc.), these differences could be weighted as well (double
processor speed doubles the impact and so forth). [0125] 6. Power
usage differences (more or less efficient hardware). [0126] 7. List
generation processing time itself.
[0127] To perform DPO score predictions for the move list
permutations, the SMM may first take into account eliminating the
wasted process of scheduling moves to systems that will cause a
violation of the requirements for some virtual machines. To do
this, a `no-move`list may be generated at this point and referred
to at various times during the process. This no-move list could
account for virtual machines that can't share the same server. For
example, a high value production virtual machine and a development
virtual machine, to lessen the chance of system failures affecting
the high value virtual machine, etc. The no-move list would be
referred to each time a potential target system is examined for a
virtual machine to move to.
[0128] To perform DPO score predictions for the move list
permutations, the SMM may account for processor and memory
licensing availability and additional costs or savings by adding or
removing resource usage in each system in the pool. Mobile capacity
on demand (CoD) is a function that helps with temporary transition
of licenses across systems for migration ease, so that can be taken
into consideration. The SMM needs to know the licensing landscape
so first of all to not exceed any licenses but also to determine if
there are cost-benefit differences between servers.
[0129] The SMM 350 performs DPO-score predictions for the move
lists on target systems. If virtual machines have the same
priority, the SMM may run DPO-score predictions on each possible
ordering of moving the virtual machines since doing DPO in a
particular order affects the overall ability to DPO scoring and the
overall scores. In other words if two virtual machines have
different configurations, but have the same priority, doing a DPO
for one virtual machine before doing DPO for the other one affects
the resulting system affinity/config layout in one way, but
performing them in reverse order will result in a different
resulting affinity/layout. So the order in which migrating the
pardons is done is important for the optimal move list.
[0130] The SMM must make a determination of how many permutations
to create and consider for move lists. The number and type of
permutations run can vary based on user input and/or configuration
factors. For example, for small numbers of virtual machines,
perhaps it's appropriate to examine all possible variations and
create all possible move order lists. If compute time is a factor,
due to the number of permutations, then limits such as always
moving the virtual machines in priority order, but the list varies
by which targets are chosen (in other words each list in this case
always has the same move order, but they vary by which target
system is affected for each virtual machine).
[0131] The SMM 350 may account for virtual machines that are
sufficiently similarly configured with like requirements and skip a
number of permutation computations if there would be no significant
difference in the resulting scores. In other words, in order to cut
down on the number of permutation calculations the SMM should
determine which virtual machines have similar characteristics and
if there are any matching ones, then just run one set of
permutations (if virtual machine A is similar enough to virtual
machine C, then there is no need to run permutations A, B, C and C,
B, A).
[0132] FIGS. 7 and 8 described in the following paragraphs
illustrate an example of creating an optimal move list for placing
virtual machines in a cloud environment based on optimized
processor-memory affinity. In this example a cloud environment
exists with eleven virtual machines spread across five servers.
Five of the virtual machines have unacceptable performance
characteristics primarily due to poor processor-memory affinity (in
this example) and are good candidates to determine if a local DPO
operation will provide enough improvement--but failing that, the
focus of this invention, computation is done to determine if
migrations to target systems, and possibly further DPO operations
on source and target systems, allow the performance targets to be
met. If so, as part of these calculations, migration move order
lists are generated with specific targets and possible DPO
operations and projected improvements are presented so that the
user is assured the best possible performance outcome when
confronted with a complicated cloud environment.
[0133] For simplicity, the example in FIGS. 7 and 8 primarily
focuses on the current and projected processor-memory affinity
scores for each migrated, and DPO-affected non-migrated, virtual
machine as well as total system scores. Also all systems are
assumed to be allowed as targets. Other variables could be taken
into account as primary or secondary factors when calculating and
performing processor-memory improvement operations. These variables
could include: servers in the cloud that contain virtual machines,
currently allocated processor counts for each server, currently
allocated memory sizes for each server, maximum processor licensing
for each server and per unit cost, maximum memory licensing for
each server and per unit cost, processor performance and power
characteristics for each server, memory performance and power
characteristics for each server, virtual machines on the servers
that need performance improvement, virtual machines on the servers
that have acceptable performance, minimum and current numbers of
processors required by each virtual machine, minimum and current
memory sizes required by each virtual machine, minimum
processor-memory affinity scores for each virtual machine and
server, known virtual machine priority values, virtual machine
security requirements, maximum migration move list calculation time
(if any), and maximum migration process and DPO process time (if
any).
[0134] FIG. 7 illustrates the initial state of five servers for
this example. Each server 710, 712, 714, 716, and 718 has the
processor, memory and affinity scores as shown. The affinity score
720 of Server1 710 is shown as ( 90/70). These numbers represent
the minimum/current affinity scores for Server1 710. The other
servers have similar affinity scores. The servers also have one or
more virtual machines as shown. For example, Server1 710 has a
virtual machine VM101 722. In this example, the first number in the
virtual machine number (i.e. the first "1" in "101) of the virtual
machine name represents that the server originated in Server1 710.
The initial conditions of the servers 710, 712, 714, 716, 718 can
be summarized as follows. For Server1 710 the system affinity score
is below its minimum goal score of 90, the server has some
processors and memory available for immediate use for possible
local DPO operations or for migration target use and both VM101 and
VM102 are below their affinity goal scores. For Server2 712, the
server affinity score is a perfect 100, with `room` to regress and
still meet the minimum goal score of 50. The server has some
processors and memory available for immediate use for possible
local DPO operations or for migration target use. The virtual
machine scores are also perfect as expected if system score is
perfect, with minimum goal scores well below the max. For Server3
714, the server affinity score 70 is below its minimum goal score
of 90. The system has some processors and memory available for
immediate use for possible local DPO operations or for migration
target use, and is below its affinity goal score. For Server4 716
the server affinity score 90 is just above the minimum goal score
of 85, the system has some processors and memory available for
immediate use for possible local DPO operations or for migration
target use and is below its affinity goal score. For Server5 718,
the server affinity score is a perfect 100 and the system does not
have any available processors or memory for DPO or target
migration.
[0135] The first step for the example shown in FIGS. 7 and 8 is to
determine the initial parameters and inputs. The initial inputs
required by the SMM in this example include: [0136] 1. The source
systems S1, S2, S3, S4, and S5 where the `S` signifies source and
the number refers to the server number 1 through 5 710, 712, 714,
716, 718. [0137] 2. The target systems T1, T2, T3, T4, and T5 (`T`
signifies target, in this example all systems can be a source
and/or target so S1 and T1 are the same system). [0138] 3. Logical
machine rankings, with VM101, VM102, VM301, VM401 and VM501 given
the highest priority. [0139] 4. Minimal affinity scores for virtual
machines. [0140] 5. Minimal affinity scores for servers are also
given, this can be an important factor if one server is more
expensive to purchase and maintain compared to another, and as such
it makes sense to utilize it to more efficiently by specifying a
higher minimal affinity score. [0141] 6. Goal affinity scores for
virtual machines. [0142] 7. Goal affinity scores for servers are
also given. [0143] 8. In this example security concerns are not
specified for simplicity. [0144] 9. In this example maximum
migration and processing times are not specified for
simplicity.
[0145] After determining initial parameters, the SMM gathers source
system information. The source system and virtual machine hardware
configurations and topology are known. Source systems S1 through S5
have virtual machines VM101, VM102, VM302, and VM403 that have
current scores below the minimum and are candidates for DPO and
migration to improve the scores. VM101 and VM102 have the highest
priority. VM302 is next in priority, and VM403 has the lowest
priority.
[0146] After gathering source system information, the SMM gathers
target system information. Target system configurations and
topology are known. In this example all systems can be used as
potential targets. Virtual machines whose affinity score is
acceptable may still be affected by DPO operations on their host
server, however the SMM should guard against the scores dropping
below the minimum score unless it is to enable a score increase to
a higher priority virtual machine with enough overall gain to
compensate (for example if the system score increases as a result
of lower priority decrease and higher priority increase). For
example, the smart migration mechanism may migrate a virtual
machine to a system that causes a lower priority virtual machine to
drop below a minimum score but the migration also causes the higher
priority virtual machine a significant increase in score such that
the migration results in an overall increase in system
performance.
[0147] After gathering source and target system information, the
SMM performs initial checks. Some virtual machines and systems have
affinity scores below their goal target, but for this example it is
determined that simple local DPO will not provide enough
improvement and that there are enough virtual machines affected
that list generation is required. The SMM then determines if a move
list is needed. Some virtual machines and systems require affinity
improvement and initial checks determined that local DPO and simple
migration is not sufficient so proceed to the generation of move
lists. In this example, we determine a move list is needed because
doing just local DPO will not be sufficient to meet the VM
requirements. Thus no local DPO operations will be done or obvious
migrations (pre-list generation improvements). In other examples,
pre-processing may be done as initial quick checks determine some
obvious simple improvements, and then new source and target
configuration information will be gathered before proceeding to
list generation.
[0148] After determining a move list is needed, the SMM generates
and ranks an optimal move list according to the following steps.
First, the SMM creates migration permutations. In this example,
there are five source systems and five target systems, with four
virtual machines that need affinity score improvement. Move list
generation first requires the calculation of the maximum number of
permutations given the number of target systems and the number of
virtual machines to migrate. In this example, even though there are
only four virtual machines to migrate, there are many theoretical
permutations. However many can be eliminated immediately, since
they are not possible (not enough resources available on the
target, security restrictions, etc.). Also a virtual machine cannot
migrate to its source server obviously, although in some
implementations it may make sense to migrate to a different target,
then eventually migrate back to the original source system after
other changes have taken affect. For example, a complicated
situation where a virtual machine is temporarily migrated off a
source system, along with possibly other `permanent` migrations off
the source system, then DPO operations are performed on the source
system that was not possible before (more resources freed up to
move), then it may arise that the temporarily migrated virtual
machine is migrated back to the original source system. It then has
the best possible affinity score that was only made possible by
`shuffling` it and other virtual machines around in the cloud. This
type of multiple target move scenario is not discussed in detail in
this disclosure however it is a natural progression of the concept
of creating move lists to single target servers.
[0149] For move list permutations, there are 1*2*3*4=24
combinations of ordering of the four virtual machines of this
example. The 24 combinations are as follows, where "Tx" means a
target system as discussed further below. [0150] 1) VM101, VM102,
VM302, VM403.fwdarw.Tx [0151] 2) VM101, VM302, VM403,
VM102.fwdarw.Tx [0152] 3) VM101, VM403, VM102, VM302.fwdarw.Tx
[0153] 4) VM101, VM102, VM403, VM302.fwdarw.Tx [0154] 5) VM101,
VM302, VM102, VM403.fwdarw.Tx [0155] 6) VM101, VM403, VM302,
VM102.fwdarw.Tx [0156] 7) VM102, VM101, VM302, VM403.fwdarw.Tx
[0157] 8) VM102, VM302, VM403, VM101.fwdarw.Tx [0158] 9) VM102,
VM403, VM101, VM302.fwdarw.Tx [0159] 10) VM102, VM101, VM403,
VM302.fwdarw.Tx [0160] 11) VM102, VM302, VM101, VM403.fwdarw.Tx
[0161] 12) VM102, VM403, VM302, VM101.fwdarw.Tx [0162] 13) VM302,
VM102, VM101, VM403.fwdarw.Tx [0163] 14) VM302, VM101, VM403,
VM102.fwdarw.Tx [0164] 15) VM302, VM403, VM102, VM101.fwdarw.Tx
[0165] 16) VM302, VM102, VM403, VM101.fwdarw.Tx [0166] 17) VM302,
VM101, VM102, VM403.fwdarw.Tx [0167] 18) VM302, VM403, VM101,
VM102.fwdarw.Tx [0168] 19) VM403, VM102, VM302, VM101.fwdarw.Tx
[0169] 20) VM403, VM302, VM101, VM102.fwdarw.Tx [0170] 21) VM403,
VM101, VM102, VM302.fwdarw.Tx [0171] 22) VM403, VM102, VM101,
VM302.fwdarw.Tx [0172] 23) VM403, VM302, VM102, VM101.fwdarw.Tx
[0173] 24) VM403, VM101, VM302, VM102.fwdarw.Tx
[0174] Referring to the permutations of the previous paragraph,
since there are five target systems, there are 24*5=120 simple
migrations with all VMs going to the same target server. However
all combination of target systems need to be considered. Since
target locations can repeat, in this example for instance the four
virtual machines might all go to T1 assuming they would fit. Thus,
for five target systems the number of combinations of "Tx" in the
above permutation list is: 5*5*5*5*5=3125. The theoretical maximum
move list combinations are then 24*3125=75000. However, this is
before accounting for a number of factors that should limit this
very large number to something manageable. The first factor is
impossible migrations. In this example, T5 has no resources to
spare, and furthermore its sole virtual machine has perfect
affinity so there is no incentive to alter the configuration or
reason to migrate off the system (to make room for a different
virtual machine needing improvement). In some instances, it might
make sense to actually migrate this virtual machine if it can still
achieve the same perfect affinity but overall help improve the
affinity score of the cloud. The second factor is duplicate virtual
machine to target ordering. In this example a theoretical maximum
combination count of 75000 was calculated, but in a very simple
example of duplicates virtual machines VMA, VMB migrating to
.fwdarw.target systems T1, T2 is the same (in the end) as VMB,
VMA.fwdarw.T2, T1. The virtual machines are going to the same
targets even if migrated in different order. However there may be
factors that affect whether or not these `duplicates` should be
eliminated, because each migration and DPO operation can have an
effect on future migrations and DPO operations, so in more
complicated scenarios even though the same virtual machine migrate
to the same target systems, because of ordering differences it may
affect the ability to DPO effectively or not on target systems if
done in a different migration order. The third factor to limit the
large number of permutations is configuration duplicates. Similarly
configured virtual machines and systems may be treated as
duplicates. There might be insignificant differences such that
these calculations can be eliminated.
[0175] The SMM generates the permutations as discussed above. There
are a number of valid migration lists in this scenario. For brevity
in this simplified example we will consider the following two lists
generated by the SMM as described above:
List 1
[0176] 1. VM101.fwdarw.T2 [0177] 2. VM102.fwdarw.T3 [0178] 3.
VM302.fwdarw.T4 [0179] 4. VM403.fwdarw.T2
List 2
[0179] [0180] 1. VM101.fwdarw.T2 [0181] 2. VM102.fwdarw.T3 [0182]
3. VM403.fwdarw.T2 [0183] 4. VM302.fwdarw.T4
[0184] The difference between the two lists for the simple example
above is the order between VM302 and VM403 migrations. VM302 is the
largest virtual machine, and by migrating VM403 first off of server
4 (which, in this list, is the target for VM302), it makes more
room for VM302 to get favorable affinity with or without DPO
operations on T4. In general, the more free resources relative to
the size of the virtual machine being moved, the greater the
possibility of achieving the affinity goal for VM403 (all other
virtual machine scores should be the same). Another related
alternative list would have the orders of VM101 and VM302 swapped,
since moving virtual machines to and virtual machines from a
particular server can affect the DPO score of other virtual
machines that are migrated to that server (so swapping VM101 and
VM302 migration order may have an impact on the DPO scores on
target server 2).
[0185] The SMM 350 then calculates target system DPO score
predictions for each of move lists. For list 1 and list 2 potential
affinity scores are calculated for the virtual machines and the
servers (if desired). In this simplified example where the
comparison is between only two lists, list 2 has a potential for
greater improvement. So, for example, potential scores are
calculated as follows:
TABLE-US-00001 List 1 VM Minimum Current Potential VM101 90 50 90
VM102 90 80 100 VM302 90 50 75 VM403 70 50 60
TABLE-US-00002 List 2 VM Minimum Current Potential VM101 90 50 90
VM102 90 80 100 VM403 70 50 60 VM302 90 50 100
[0186] VM302 not only should achieve its goal with the second list,
it should exceed it. Note that VM403 did not achieve its goal with
either list, however it's still an improvement, shown here to
demonstrate that it might not be possible to always achieve the
affinity goal. Further, server 4 system affinity score should be
higher than in the current state, possibly for list 1 (depending on
if the removal of VM403 and the addition of VM302 produces a better
system score), but certainly for list 2 since in that case an
under-performing virtual machine was exchanged with a virtual
machine with an above goal score.
[0187] After calculating target system DPO score predictions, the
SMM produces ranked migration lists. In the above example lists,
list 2 is ranked above list 1 based on a couple of simplified
factors. All else being equal, VM302 affinity score is better with
list 2 than with list 1, and server 4's affinity score should be
higher with list 2 than with list 1.
[0188] After determining ranked lists, the SMM 350 may allow a user
or system administrator to change the initial parameters as
described above. The administrator may wish to alter parameters
such as virtual machine priorities, affinity goals, available
target systems etc. The SMM may allow the user to do these things
before proceeding with the migrations. If anything changed, then
the number and nature of the permutations may need to be changed,
or perhaps the list rankings or server scores need to be altered.
If for example, virtual machine priorities are changed, then that
would affect server affinity scores and list rankings. The
administrator may have the option to choose among lists depending
on which factors are important. The administrator may just consider
the best affinity scores for the highest ranked virtual machines,
or perhaps the best affinity scores of the servers in the cloud, or
perhaps some balance between score improvement and less disruption
etc. The administrator (or automation) then selects the best list
to proceed. The list would denote the order of the migrations of
course, but also may indicate at which step DPO operations are to
be performed (on source and/or target systems before or after
virtual machine X moves from and to server Y, etc). The SMM 350
then starts migrations and any DPO operations. In this example, we
assume the administrator chooses to migrate virtual machines
according to list 2. After the migrations, the virtual machines
appear as shown in FIG. 8 with the improved affinity scores as
shown. The SMM 350 may monitor the migration and DPO operations to
ensure the processing does not exceed a pre-selected maximum time.
The SMM 350 and DPO 508 may keep track of actual vs predicted
affinity scores to present to the administrator as the process
continues to completion.
[0189] FIG. 9 illustrates a flow diagram of a method 900 for
placing virtual machines in a cloud environment based on optimized
processor-memory affinity. The method 900 is presented as a series
of steps performed by a computer software program such as the smart
migration mechanism 350 described above. First, determine initial
parameters (step 910). Gather affinity scores and configuration
data from source systems (step 920). Gather affinity scores,
configuration data and resources available from target systems
(step 930). Perform initial checks and determine if requirements
can be met without DPO operations and move lists (step 940). If
move lists are not needed (step 950=no) then perform DPO and simple
migrations as needed (step 960) and the method is then done. If
move lists are needed (step 950=yes) then create an optimal move
list using DPO-score predictions on the target systems (step 970).
Migrate virtual machines from the source system to the target
systems using the optimal move list (step 980). The method is then
done.
[0190] Referring now to FIG. 10, a flow diagram shows method 1000
that is an exemplary method for performing step 970 in method 900.
The method 1000 is presented as a series of steps performed by a
computer software program described above as the application smart
migration mechanism 350. First, create multiple move permutation
for move lists (step 1010). Perform DPO-score predictions of
placing virtual machines on the target systems for each list (step
1020). Give the user ranked lists of move permutations and scores
(step 1030). Allow the user to change the initial parameters (step
1040). If the parameters are changed (step 1050=yes) then go to
step 1010. If the parameters are not changed (step 1050=no) then
allow the user to select a move list (step 1060). Use the selected
move list for the optimal move list (step 1070). The method is then
done.
[0191] Referring now to FIG. 11, a flow diagram shows method 1100
that is an exemplary method for performing step 1020 in method
1000. The method 1100 is presented as a series of steps performed
by a computer software program described above as the application
smart migration mechanism 350. First, create a no-move list (step
1110). Generate a number of possible move lists (step 1120).
Perform DPO-score predictions on the move lists for the target
systems by scoring primary and secondary factors (step 1130). The
method is then done.
[0192] The claims and disclosure herein provide an apparatus and
method for placing virtual machines in a cloud environment based on
optimized processor-memory affinity. A smart migration mechanism
(SMM) predicts an optimization score for multiple permutations of
placing virtual machines on a target system to create an optimal
move list to allow the user to select an optimal move list to
migrate the virtual machines from the source system to enhance the
performance of the overall systems.
[0193] One skilled in the art will appreciate that many variations
are possible within the scope of the claims. Thus, while the
disclosure is particularly shown and described above, it will be
understood by those skilled in the art that these and other changes
in form and details may be made therein without departing from the
spirit and scope of the claims.
* * * * *