U.S. patent number RE37,494 [Application Number 08/122,934] was granted by the patent office on 2002-01-01 for transfer of messages in a multiplexed system.
This patent grant is currently assigned to QPSX Communications Pty. Ltd.. Invention is credited to Antonio Cantoni, Robert M. Newman.
United States Patent |
RE37,494 |
Cantoni , et al. |
January 1, 2002 |
**Please see images for:
( Certificate of Correction ) ** |
Transfer of messages in a multiplexed system
Abstract
A method and apparatus for transmitting variable length messages
on a network in fixed length slots including the provision of a
source identifier field in the header of each slot, the source
identifier field including a code which is uniquely associated with
the message to be transmitted. The fixed length slots are then
transmitted on the network and reassembly of the slots by a
reassembly machine is controlled in accordance with the identifier
codes in the slot.
Inventors: |
Cantoni; Antonio (North Perth,
AU), Newman; Robert M. (Santa Clara, CA) |
Assignee: |
QPSX Communications Pty. Ltd.
(AU)
|
Family
ID: |
3772066 |
Appl.
No.: |
08/122,934 |
Filed: |
September 17, 1993 |
PCT
Filed: |
March 17, 1988 |
PCT No.: |
PCT/AU88/00075 |
371
Date: |
April 28, 1989 |
102(e)
Date: |
April 28, 1989 |
PCT
Pub. No.: |
WO88/07293 |
PCT
Pub. Date: |
September 22, 1988 |
Related U.S. Patent Documents
|
|
|
|
|
|
|
Application
Number |
Filing Date |
Patent Number |
Issue Date |
|
Reissue of: |
283364 |
Apr 28, 1989 |
05050166 |
Sep 17, 1991 |
|
|
Foreign Application Priority Data
Current U.S.
Class: |
370/395.6;
340/7.2; 370/471; 370/474 |
Current CPC
Class: |
H04J
3/16 (20130101); H04L 29/06 (20130101); H04L
47/10 (20130101); H04L 47/34 (20130101); H04L
49/606 (20130101); H04L 2012/565 (20130101); H04L
2012/5652 (20130101); H04L 2012/641 (20130101); H04L
69/324 (20130101) |
Current International
Class: |
H04L
12/56 (20060101); H04L 12/64 (20060101); H04J
003/24 (); H04Q 011/04 (); H04L 012/54 () |
Field of
Search: |
;340/825.03,825.04,825.05,825.34,825.44,825.52
;371/55,56,67.1,68.1,71
;370/389,395,396,397,400,409,412,428,429,465,470,471,474 |
References Cited
[Referenced By]
U.S. Patent Documents
Foreign Patent Documents
|
|
|
|
|
|
|
0055674 |
|
Jul 1982 |
|
EP |
|
0079426 |
|
May 1983 |
|
EP |
|
0212701 |
|
Mar 1987 |
|
EP |
|
1326569 |
|
Aug 1973 |
|
GB |
|
1427662 |
|
Oct 1974 |
|
GB |
|
Other References
Conference Record from International Conference on Communications,
Gadre, "Comparison of multiple access protocols for packet
switching in satellite-switched multi-beamsystems," Jun. 18-20,
1980, Department of Electrical Engineering, Columbia University,
New York, N.Y.; 8 pp. .
Conference literatue from IEEE Infocom '86, Pavey et al., "A
Performance Evaluation of the PDAMA Satellite Access Protocol" Apr.
8-10, 1986, E-Systems, Inc., ECI Division, St. Petersburg, FL; 12
pp. .
Literature from IEEE International Conference on Communications
'88, Wong et a;., "The Controlled-SRMA Protocol for Packet
Satellite Communication", Jun. 12-15, 1988, Department of
Electronics, The Chinese University of Hong Kong, Shatin, Hong
Kong, 9 pp. .
Literature, Chitre, "Capacity Allocation Scheme for Transmission of
Packerts Over Satelitte Links", 1983, Communications Satellite
Corporation 1983, 20 pp. .
Literature, Benelli et al., "Integration of Random Access and Time
Division Techniques in Satellite Communications" Universita Di
Firenze, Italie; 5 pp. .
Article, Ilyas et al., "Effects of Message Segmentation in
Tadem-Mode Computer Networks" 1987, Computer Engineering, florida
Atlantic University, Boca Raton, FL; 5 pp. .
Literature, Limb et al., "Improved scheduling of traffic for a
high-speedslooted ring," Electronics Letters Jul. 16, 1987, vol.
23, No. 15, 3 pp. .
Literature, Kobayashi et al., "Satellite Packet Communication
Architeture" C&C Systems Research Labs, NEC Corporation, 5 pp.
.
Literature, Flak et al., "A Multiprocessor Channel Scheduler for
the Wideband Packet Satellite Network" 1983, Bolt Beranek and
Newman Inc., Cambridge, MA, 8 pp. .
Literature, Tasaka et al., "Performance Analysis of the SRUC
Protocol with a Go-Back-N ARZ Scheme for Satellite Broadcast
Channels" 1984, Department of Information Engineering, Nagoya
Institute of Technology, Nagoya, Japan, 6 pp. .
Literature, Takizuka et al., "Design and Evaluation of Satellite
Packet Communication Protocols for Integrated Services Networks"
1983, KDD, Research and Development Laboratories, 6 pp. .
Article, Li, "Multiple Access Communications Networks" IEEE
Communications Magazine, Jun. 1987, vol. 25, No. 6, 8 pp. .
Literature, Bolus, Fallour et al., "Computer Protocols and
Applications Using Satellite Networks: THE NADIR Project" Computer
Science Press, Inc., 18 pp. .
Literature, Chlamtac et al., "Performance of Multibeam Packet
Satellite System with Conflict Free Scheduling" 1986, Department of
Computer Sceince, Technion-Israel Institute of Technology, Haifa,
Israel, 8 pp. .
Literature, Koubias et al., "A New Mulitple Access Protocol and its
Simulated Performance" Applied Electronic Laboratory, School of
Engineering, University of Patras, Greece, 7 pp. .
Literature, Lam, "Protocols for Satellite Packet Switching"
Department of Computer Sciences, The Unievrstiy of Texas at Austin,
Austin, Texas, 6 pp. .
Literature, Jacobs et al., "Packet Satellite Network Design Issues"
1979, LINKABIT Corporation, San Diego, CA, 12 pp. .
Literature, Mukherjee et al., "Multichannel Dynamic Satellite
Packet Networkd for Computer Databanks" 1983, Southern Illinois
University, Carbondale, 3 pp..
|
Primary Examiner: Hsu; Alpus H.
Attorney, Agent or Firm: Fish & Richardson P.C.
Claims
The claims defining the invention are as follows:
1. A method of transmitting variable length messages on a network
from a source having a source address to a destination having a
destination address, said method including the steps of:
segmenting each message into a plurality of fixed length slots
including a first slot, continuing slots, and a last slot,
.[.each.]. of said slots including a header field which includes a
source identifier field, which is substantially shorter than said
destination address, and a message segment;
providing a source identifier code in the source identifier field,
each source identifier code being uniquely associated with the
message to be transmitted;
entering said destination address in the message segment of said
first slot;
transmitting the slots on the network; and
controlling reassembly of slots at the destination in accordance
with the source identifier code of slots received at the
destination.
2. A method as claimed in claim 1, including the step of providing
.[.the .Iadd.a.Iaddend. type field in the header field of each
slot, and coding into the type field a first, second or third code
representing a beginning of message, a continuation of message and
an end of message respectively and controlling the reassembly of
received slots at the destination in accordance with the first,
second and third code.
3. A method as claimed in claim 2, wherein the method includes the
step of storing message segments associated with a single message
in a buffer at the destination.
4. A method as claimed in claim 3, wherein if said first code is
detected at the destination, the source identifier code is inputted
to a comparator and if said second code associated with a
subsequently received slot is detected the source identifier
thereof is also inputted to the comparator to check for a match,
and if a match occurs the message segment of the subsequently
received slot is stored in said buffer.
5. A method as claimed in claim 4, wherein if said third code is
detected a reassembled message in the buffer is outputted from the
buffer.
6. A method as claimed in claim 2, including the step of coding
into the type field a fourth code representing a single segment
message and if said fourth code is detected in a slot received at
the destination, the message segment thereof is stored in a single
segment buffer.
7. A method as claimed in claim 5, including the step of providing
multiple comparators and buffers at the destination so as to enable
simultaneous receipt of a plurality of messages each having its own
source identifier code, the message segments of each message being
stored in a single buffer.
8. A method as claimed in claim 1, including the step of
concurrently transmitting two or more messages from the source to
the destination on the network.
9. Apparatus for transmitting variable length messages on a network
from a source having a source address to a destination having a
destination address in fixed length slots, said apparatus
including:
a segmentation machine for segmenting each message into a plurality
of fixed length slots including a first slot, continuing slots, and
a last slot, each of said slots including a header field which
includes a source identifier field which is substantially shorter
than said destination address, and a message segment,
coding means for providing a source identifier field including a
source identifier code which is uniquely associated with the
message to be transmitted,
means for entering said destination address in the message segment
of said first slot, and
a reassembly machine located, in use, at the destination, said
reassembly machine including control means for controlling
reassembly of slots in accordance with the source identifier codes
of the slots.
10. Apparatus as claimed in claim 9, wherein said coding means
provides a type field in the header field of each slot and provides
a first, second or third code representing a beginning of message,
a continuation of message and an end of message, respectively, and
wherein the control means is responsive to said first, second and
third codes.
11. Apparatus as claimed in claim 10, wherein said reassembly
machine includes detecting means for detecting in the header fields
of the received slots the presence of the third code and for
detecting a match between the destination address in the slot and
the destination address of the destination, and wherein if there is
an address match, the detecting means copies the source identifier
code into a comparator means.
12. Apparatus as claimed in claim 11, wherein the reassembly
machine includes a plurality of said comparator means to enable
concurrent receipt of slots of different messages.
13. Apparatus as claimed in claim 12, including means inputting the
source identifier codes of received slots to said plurality of
comparator means to thereby enable matching of slots having the
same source identifier codes.
14. Apparatus as claimed in claim 13, including a plurality of
buffers for the message segments of the sots and a buffer selector
circuit which selects a particular buffer for receipt of all
message segments of slots having the same source identifier code
whereby reassembled messages are stored in said buffers..Iadd.
15. A method of transmitting variable length messages on a network
from a source to a destination, said method including the steps of
segmenting each message into a plurality of fixed length slots each
of which slots includes a header field and a message segment,
providing a source identifier field in the header field of each
slot, said source identifier field including a source identifier
code that is uniquely associated with the message to be
transmitted, transmitting the slots on the network, controlling the
reassembly of slots at the destination in accordance with the
source identifier codes of the slots received at the destination,
and providing a type field in the header field of each slot, and
coding into the type field a code selected from a first code, a
second code and a third code, respectively representing a beginning
of message, a continuation of message or an end of message, and
controlling the reassembly of received slots at the destination in
accordance with said source identifier codes and the first code,
the second code and the third code..Iaddend..Iadd.
16. A method as claimed in claim 15 wherein the message includes a
destination address field that is checked for a match with an
address associated with the destination, and wherein the address
field is transmitted in the message segment of the first slot used
to transmit the message..Iaddend..Iadd.
17. A method as claimed in claim 15 or 16 wherein the method
includes the step of storing message segments associated with a
single message in a buffer..Iaddend..Iadd.
18. A method as claimed in claim 17 wherein if said first code is
detected at the destination, the source identifier code is inputted
to a comparator and if the second code associated with a
subsequently received slot is detected the source identifier
thereof is also inputted to the comparator to check for a match,
and if a match occurs the message segment of the subsequently
received slot is stored in said buffer..Iaddend..Iadd.
19. A method as claimed in claim 18 wherein if said third code is
detected the reassembled slots in the buffer are outputted from the
buffer as a reassembled message..Iaddend..Iadd.
20. A method as claimed in claim 19 including the step of providing
multiple comparators and buffers at the destination so as to enable
simultaneous receipt of a plurality of messages each having its own
source identifier code, the message segments of each message being
stored in respective buffers..Iaddend..Iadd.
21. A method as claimed in claim 15 including the step of coding
into the type field a fourth code representing a single segment
message and if said fourth code is detected in a slot received at
the destination, the message segment thereof is stored in a single
segment buffer..Iaddend..Iadd.
22. Apparatus for transmitting variable length messages on a
network from a source to a destination in fixed length slots said
apparatus including;
a segmentation machine for segmenting the messages into fixed
length slots, each of said slots includes a header field and a
message segment, said segmentation machine including coding means
for providing a source identifier field in the header of each slot,
the source identifier field being substantially shorter than any
associated destination address, said source identifier field
including a source identifier code that is uniquely associated with
the message to be transmitted, and a reassembly machine located, in
use, at the destination, said reassembly machine including control
means for controlling reassembly of slot in accordance with the
source identifier codes of the slots..Iaddend..Iadd.
23. Apparatus as claimed in claim 22 wherein said coding means
provides a type field in the header field of each slot and provides
a code selected from a first code, a second code and a third code,
respectively representing a beginning of message, a continuation of
message and an end of message, and wherein the control means is
responsive to said source identifier codes and said first code,
said second code and said third code..Iaddend..Iadd.
24. Apparatus as claimed in claim 22 or 23 wherein the message
includes a destination address field and wherein the segmentation
machine is arranged to transmit the destination address field in
the message segment of the first slot of the
message..Iaddend..Iadd.
25. A method of transmitting variable length messages on a network
from a source having a source address to a destination having a
destination address, said method including the steps of:
segmenting each message into a plurality of fixed length slots
including a first slot, continuing slots, and a last slot, each of
said slots including a header field that includes a source
identifier field, which is substantially shorter than said
destination address, and a message segment;
providing a source identifier code in the source identifier field,
said source identifier code being uniquely associated with the
message to be transmitted;
transmitting the slots on the network; and
controlling reassembly of slots at the destination in accordance
with the source identifier codes of the slots received at the
destination..Iaddend..Iadd.
26. Apparatus for transmitting variable length messages on a
network from a source having a source address to a destination
having a destination address in fixed length slots, said apparatus
including:
a segmentation machine for segmenting each message into a plurality
of fixed length slots including a first slot, continuing slots, and
a last slot, each of said slots including a header field that
includes a source identifier field that is substantially shorter
than said destination address, and a message segment;
coding means for providing the source identifier field with a
source identifier code that is uniquely associated with the message
to be transmitted; and
a reassembly machine located, in use, at the destination, said
reassembly machine including control means for controlling
reassembly of slots in accordance with the source identifier codes
of the slots..Iaddend..Iadd.
27. A method for the connection oriented transfer of variable
length messages (20) on a network having a plurality of nodes (4)
from a source node (42) having a source address (24) to a
destination node (46) having a destination address (22), said
method including the steps of:
segmenting each message (20) into a plurality of fixed length slots
(32) including a first slot, continuing slots and a last slot, each
of said slots including a header field (34, 36, 38) and a message
segment (40);
transmitting the slots from the source node to the network; and
controlling reassembly of slots received at the destination node
into the message on the basis of information in the header
field;
a source identifier code (SI) uniquely associated with the message
to be transmitted from the source node being provided in a source
identifier field in the header field of each of said slots; and
said reassembly of slots at the destination node being controlled
in accordance with the source identifier code of slots received at
the destination node..Iaddend..Iadd.
28. A method as claimed in claim 27 including the step of providing
a type field (36) in the header field of each slot, and coding into
the type field a code representing an end of message (EOM) and
controlling the reassembly of received slots (32) at the
destination node in accordance with said code..Iaddend..Iadd.
29. A method as claimed in claim 28 including the step of storing
message segments (40) associated with a single message in a buffer
(77) at the destination node..Iaddend..Iadd.
30. A method as claimed in claim 29 wherein the source identifier
code (SI) of the first slot received at the destination node is
inputted to a comparator (82) and the source identifier code (SI)
of each subsequently received slot is also inputted to the
comparator to check for a match, and if a match occurs, the message
segment (40) of the subsequently received slot is stored in said
buffer (77)..Iaddend..Iadd.
31. A method as claimed in claim 30 wherein if one of said codes is
detected, the reassembled message (20) in the buffer (77) is
outputted from the buffer (77)..Iaddend..Iadd.
32. A method as claimed in claim 30 or 31 including the step of
providing multiple comparators (82) and buffers (77) at the
destination node so as to enable simultaneous receipt of a
plurality of messages (20) each having its own source identifier
code (SI), the message segments (40) of each message (20) being
stored in a single buffer (77)..Iaddend..Iadd.
33. Apparatus for the connection oriented transfer of variable
length messages (20) on a network having a plurality of nodes (4)
from a source node (42) having a source address to a destination
node (46) having a destination address in fixed length slots (32),
said apparatus including:
a segmentation machine (48) located, in use, at the source node,
for segmenting each message (20) into a plurality of fixed length
slots (32) including a first slot, continuing slots and a last
slot, each of said slots including a header field (34, 36, 38) and
a message segment; and
a reassembly machine (50) located, in use, at the destination node,
said reassembly machine including control means for controlling
reassembly of slots (32) in accordance with information in the
header field;
coding means for providing a source identifier field (38) in the
header field of each slot including a source identifier code (SI)
which is uniquely associated with the message (20) to be
transmitted; and
the control means being adapted to control reassembly of the slots
into the message in accordance with the source identifier codes
(SI) of the slots..Iaddend..Iadd.
34. Apparatus as claimed in claim 33, wherein said coding means
provides a type field (36) in the header field of each slot and
provides a further code representing an end of message (EOM) and
wherein the control means is operative to control reassembly of the
slots to a reassembled message..Iaddend..Iadd.
35. Apparatus as claimed in claim 34 wherein the control means
includes selector means (70) for checking said further code in the
header field of the slots and for inputting the source identifier
codes (SI) to comparator means (82) for checking for matches with
subsequently received source identifier codes..Iaddend..Iadd.
36. Apparatus as claimed in claim 35, wherein the reassembly
machine (50) includes a plurality of said comparator means (82) to
enable concurrent receipt of slots of different
messages..Iaddend..Iadd.
37. Apparatus as claimed in claim 36, including means for inputting
the source identifier codes (SI) of received slots to said
plurality of comparator means (82) to thereby enable matching of
slots having the same source identifier codes
(SI)..Iaddend..Iadd.
38. Apparatus as claimed in claim 37, including a plurality of
buffers (77) for the message segments of the slots and a buffer
selector circuit which selects a particular buffer for receipt of
all message segments of slots having the same source identifier
code (SI) whereby reassembled messages are stored in said
buffers..Iaddend..Iadd.
39. Apparatus as claimed in claim 35 wherein, if the selector means
(70) detects a further code representing the end of a message
(EOM), the control means operates to output a reassembled message
from the buffer (77) which is associated with the source identifier
code (SI) of the slot containing said detected further
code..Iaddend.
Description
This invention relates to a transfer of messages in a multiplexed
system.
More particularly, the invention relates to a method for the
transfer of messages in a time multiplexed slotted environment such
as a communications network. The network can be of the type
disclosed in International Publication No. WO 86/03639 and that
disclosure is incorporated herein by cross-reference.
Generally speaking, the invention provides an efficient method for
the connectionless or connection oriented transfer of messages of
arbitrary but finite length in a time multiplexed slotted
environment with constrained destination resources.
The method provides for the efficient support of any type of
addressing (short or extended, hierarchical or non-hierarchical) in
the one environment, even in a system with short slots.
In the one environment, the method can guarantee delivery of
messages or provide a more efficient transfer at the expense of
occasional message loss. In either case, there are minimal
communication overheads and the utilisation of destination
resources is maximised. Thus the method provides a wide range of
options and considerable scope for achieving a range of
performance-cost objectives.
BACKGROUND ART
In the data communications environment, information is generally
exchanged in units called packets. These consist of an overhead
necessary for the control and addressing of the unit through the
data switch and of the actual information. Typically the size of
the information unit is not fixed but depends upon the message and
the amount of information to be transferred.
Early packet switches handled the variable length packets as a
whole unit, allocating all its communication resource to the
transfer of the packet until its completion. There are a number of
new packet switch designs emerging that switch only small fixed
length slots. These switches are commonly referred to as Fast
Packet Switches. Such switches are an improvement since they are
generally simpler, may operate at higher speeds and allow for the
support of real-time traffic.
If the Fast Packet Switch is to carry packet communications of
variable length then it is necessary that the original packet
message be segmented for transmission over the switch and
reassembled at the destination. The segmentation function is
relatively simple only requiring that the message be divided into
units of size equal to or smaller than the slot size. The
transmission of the segments over the packet switch however
requires much more since it is necessary that the destination can
receive and order all of the segments of the message. Thus it is
required that there be a logical association between all of the
slots of a single message. The reassembly function then
reconstructs the original variable length message from all of the
received segments. Special care may need to be taken in the
reassembly function to allow for the possibility that more than one
message may need to be reassembled concurrently.
Some schemes have been developed to perform the segmentation and
reassembly function. However, these are limited either in the
efficiency of the transport in the switch or in the performance of
the reassembly function. The efficiency of the transport is
typically limited by the overheads that are carried on each slot.
These overheads are required to route the slot to the destination
and to control the reassembly of the message at the destination. An
example of this is the Slotted Ring protocol where 13 bytes of
overhead are required in each slot as described in I.E.E.E. 802.6
Draft Standard Proposal "Slotted Ring" Sept. 1986. The overheads in
that case include addressing, sequence indication, and length
indication.
The problem with the slot overhead is compounded by addressing
requirements. The common address field sizes used in data
communications are 16 to 48 bits. With 48 bit addressing there is
an overhead of 12 bytes per slot (source and destination address)
in addition to the reassembly overheads. This approach to
segmentation is clearly inefficient with small (less than 32 bytes)
slot sizes.
The overhead problem can be reduced by logically associating the
segments of the same message by the use of a count scheme, as
suggested in an article by K. Yukimatsu, N. Watanabe, T. Honda
"Multicast Communication Facilities in a High Speed Packet
Switching Network", Proc. ICCC 86 Munich Sept. 1986. pp 276-281. In
this approach the segments of the message are transmitted with a
two octet i.e. 16 bits overhead, a count field. The count field
gives the number of slots separation between two consecutive
segments of the same message. By the use of this count the
destination can determine all slots of the message. The limitation
with this approach is that the number of slots between consecutive
segments of the message is limited by the maximum value of the
count field. Also, in the case of a multiple access switch, the
source cannot transmit more than one message at a time. This
reduces the efficiency of transfer when connection orientated
reassembly schemes are used.
BRIEF DESCRIPTION OF THE INVENTION
According to the present invention there is provided a method of
transmitting variable length messages on a network from a source to
a destination in fixed length slots which include a header field
and a message segment, said method including the steps of providing
a source identifier field in the header field of each slot, said
source identifier field including a source identifier code which is
uniquely associated with the message to be transmitted,
transmitting the slots on the network, and controlling the
reassembly of slots at the destination in accordance with the
source identifier codes of the slots received at the
destination.
The invention also provides an apparatus for transmitting variable
length messages on a network from a source to a destination in
fixed length slots said apparatus including:
a segmentation machine for segmenting the message into fixed length
slots which include a header field and a message segment, said
machine including coding means for providing a source identifier
field in the header of each slot, said source identifier field
including a source identifier code which is uniquely associated
with the message to be transmitted, and a reassembly machine
located, in use, at the destination, said reassembly machine
including control means for controlling reassembly of slots in
accordance with the source identifier codes of the slots.
BRIEF DESCRIPTION OF THE DRAWINGS
FIG. 1 is a diagrammatic representation showing a variable length
message and fixed length segments;
FIG. 2 is a block diagram showing the segmentation and reassembly
machines coupled to a fast packet switch;
FIG. 3 diagrammatically illustrates the segmentation and reassembly
machines coupled to a QPSX network;
FIG. 4 is a block diagram of an access unit (AU) of the network
shown in FIG. 3;
FIG. 5 is a more detailed block diagram of the access unit;
FIG. 6 is a segmentation state machine diagram;
FIG. 7 is a diagram showing request and confirm primatives for
segment transfer to the access unit;
FIG. 8 is a reassembly state machine diagram;
FIG. 9 is a diagram showing an indication primative for segment
transfer from an access unit;
FIG. 10 is another reassembly state machine diagram;
FIG. 11 is a block diagram of a reassembly machine; and FIGS. 12A
and 12B are flowcharts illustrating logical operations in the
reassembly machine.
DETAILED DESCRIPTION OF THE EMBODIMENTS OF THE INVENTION
FIG. 1 diagrammatically shows a message 20 which is of variable
length. The message includes address fields 22 and 24 for the
destination address (DA) and source address (SA). The message
includes a length field (L) 26 and an information field (IU) 28
followed by an error checking field (CRC) 30. The error checking
field 30 can be of any known type.
In accordance with the invention, the variable length message 20 is
segmented into a number of slots 32 of equal length. The method
places the address field 22 and 24 in the first of the slots 32 and
subsequently logically associates the following slots of the
message with the first slot using a unique identifier, as is
diagrammatically illustrated in FIG. 1.
Each slot 32 has an Access Control Field (ACF) 34, TYPE field 36
and Source Identifier (SI) field 38. In accordance with the
invention, the SI field 38 is used to provide the logical linking
between slots 32 of the same message. The slots 32 also include an
information field 40 which is used to carry information from the
original message 20. The ACF 34 is related to controlling
transmission of the slots through the network.
The TYPE field 36 within the slot can be used for the indication of
a wide range of information. In one use it can allow for the
separation of different communication classes within the one
environment. As an example the messages with different field sizes
and structures could be carried on the same switch. Also the TYPE
field can be used to identify between different slot structures.
Other segmentation schemes such as those described in the
Background Art could be used in the same environment and
distinguished by the TYPE field.
The TYPE field specifies how the SI field 38 and the information
field of each slot should be interpreted depending on, among
others, the following factors:
(a) Whether connectionless or connection oriented communication is
supported.
(b) Whether short or long source and destination addresses are to
be supported.
(c) Whether extended hierarchical or non hierarchical source and
destination addressing is supported.
(d) Whether the slot is carrying the first, last or continuation
segment of a message.
(e) Whether the message fits in a single segment.
(f) Whether or not the source supports message transmission to
multiple destinations concurrently.
The SI field 38 is a label which enables the logical association of
all segments 40 belonging to the one message and hence enables them
to be reassembled into the original message 20.
FIG. 2 diagrammatically illustrates a communications system
comprising a source 42 which produces messages 20 of variable
length for transmission on a network or fast packet switch 44 to a
destination 46. The system includes a segmentation machine 48
coupled between the source 42 and the switch 44 and a reassembly
machine 50 coupled between the switch 44 and the destination 46.
The segmentation machine 48 converts messages 20 of indefinite
length to slots 32 of fixed length for transmission on the switch
44. The reassembly machine 50 reassambles the slots 32 into the
original message 20 for input to the destination 46. The
segmentation and reassembly machines 48 and 50 would be located at
respective nodes or access units coupled to the network.
FIG. 3 illustrates in more detail the use of the segmentation and
reassembly machines 48 and 50 in a QPSX network of the type
disclosed in WO 86/03639. The QPSX network comprises two
unidirectional buses, bus A and bus B with data flowing in opposite
directions, a central controller 2 and a number of distributed
nodes or access units (AU's) 4 coupled between the buses A and B.
Although each bus originates and terminates at the central
controller 2 neither has a through connection, in normal
circumstances. Each AU 4 has read taps 6 and 8 from the respective
buses and lines 10 and 12 connected to unidirectional write
couplers to the respective buses. The write transmit only in the
direction of propagation of the respective buses. The read
connections for each AU are attached to the bus ahead of the write
connections and consequently the information read by each AU is
uneffected by that written by it. In the illustrated arrangement, a
source 42 is coupled to one of the access units 4 via the
segmentation machine 48. The access unit transmits the message in
fixed length slots on the network to the access unit 4 associated
with the destination 46. Normally each access unit would have both
segmentation and reassembly machines 48 and 50 to enable two way
communications. The segmentation and reassembly machines 48 and 50
can be regarded as part of the interface IP 16.
FIG. 5 shows in more detail the connection of the segmentation and
reassembly machines 48 and 50 to an access unit 4 of the type
described in the aforementioned publication. FIG. 5 corresponds
generally to FIG. 12 of that specification and hence need not be
described in more detail here.
Each node in the network will have one or more unique SI's. Each SI
can be used by the node for the transfer of a message. When the
message transmission is complete the SI can be reused. Multiple
SI's for a single node allow that node to transfer more than one
message concurrently.
To describe the operation of the message transfer scheme, the
segmentation of the message into slots is considered first and the
action at the receiver is considered after that.
The train of slots 32 sent by the segmentation machine 48 is shown
in FIG. 1. The first slot of a multisegment message will be
identified as such by a BOM (Beginning Of Message) code in the TYPE
field 36. The SI field 36 is set by reference to the unique SI of
the source node and the information field 40 contains the first
segment of the message. Thus the DA field 22 of the message 20 is
at the head of the information field 40. The following segments of
the message until the last are each placed in the information
fields of slots with the TYPE field 36 set to COM (Continuation Of
Message) and the SI field 38 containing the source's code which is
unique for this message. The last slot of a multisegment message
has the TYPE field 36 coded as EOM (End Of Message), as shown.
For the transfer of a message 20 that only requires a single slot
32 the SSM (Single Slot Message) code is used in the TYPE field 36.
The SI is not required in this case, however it is still used for
consistency in operation.
An implementation of the segmentation machine 48 will now be
described with reference to the state diagram which is shown in
FIG. 6. In this diagram, the condition for a state transisition is
shown above the transition line and the action taken is below the
line, in accordance with standard notation. This state machine will
handle the receipt of a message from a single source at a time. If
simultaneous receipt of more than one message is required, multiple
state machines would be required.
To specify the communications between adjacent parts of the system,
there are three communications primitives, as follows:
(i) Request (REQ): This is a request to send a unit of data,
(ii) Indication (IN): This is an indication that a unit of data has
been received, and
(iii) Confirmation (CONFIRM): This is a confirmation that a unit of
data has been sent without error.
Between the machines 48 and 50 and the fast packet switch 44, (via
the access units 4) communication primitives are prefixed by
SAR_ACC hence there are three communication primitives as
follows:
SAR_ACC Request
SAR_ACC Indication
SAR_ACC Confirm.
Also for source or destination equipment 42 and 46 such as a
computer attached to the segmentation and reassembly machines 48
and 50, communication primitives are prefixed by APP_SAR hence
there are three communication primitives as follows:
APP.sub.13 SAR Request:
APP_SAR Indication
APP_SAR Confirmation
A segmentation machine operates on an unconfirmed APP SAR request,
provided that sufficient resources are available. Tags for SAP_ACC
requests and Source Identifiers (SIs) are common resources for all
of these machines. Tags used in SAP_ACC requests are unique over
all such requests from any segmentation machines. A tag is
allocated on an SAR_ACC request and deallocated on an SAR_ACC
confirm. The tag for the segments are local to the particular
segment at the segmentation machine and is not transmitted on the
network. Tags are reused when message transmissions have been
completed. The tag is coded in TAG fields 52 and 54 in Request and
Confirm primitives 56 and 58. The TAG fields 52 and 54 in the
Request and Confirm primitives 56 and 58 are used for
communications between the segmentation machine 48 and access unit
4, as seen in FIG. 7. The Request primative 56 comprises the TAG
field 52, T-SEG field 60 together with the slot 32. The fields 60
and 52 are control fields which are not transmitted beyond the
access unit 4. The Confirm primative 58 comprises a TAG field 54
and TR field 64, being control fields for communication between the
access unit 4 and the segmentation machine 48. The codes in the TAG
fields 52 and 54 must correspond for a particular segment 40.
Therefore a particular message 20 will have a number of TAG codes
corresponding to the various segments 32 required to be used for
the transmission. Successful transmission or otherwise is indicated
by the coding in the TR field 64.
Source Identifiers (SI) are selected at the source so that the SI
will identify a unique message to any reassembly machine. The SI is
allocated on an APP_SAR request and is deallocated in an
implementation dependent manner.
One reasembly machine 50 is needed at the destination 46 for each
SI. A reassembly packet timer (not shown) is also associated with
each reassembly machine, to prevent locking in a wait state for
signals.
An SAR_ACC confirm is routed to the segmentation machine which
generated the SAR_ACC request. An SAR ACC indication is routed to
the reassembly machine associated with the SI. The segmentation
function of the segmentation machine 50 is described in the
following paragraphs.
1.1 ABBREVIATIONS USED IN THE SEGMENTATION STATE DIAGRAM OF FIG.
6
CONFIRMSBACK CONFIRMSBACK is an array of (TAG) flags indexed by tag
number. A CONFIRMSBACK flag is set when the SARACC confirm has heen
received corresponding to the SARACC request with tag number TAG.
ACB: All Confirms Back. Set if CONFIRMSBACK(1) is set for all
segments. CHECKSUM: Flag for checking sum of bits successfully
transmitted. LAST SEGMENT SENT: Flag indicating whether all
segments for this message has been sent. LAST SEGMENT SENT: Flag
indicating that not all segments for this message has been sent.
SI: The source identifier from the segment in an SARACC indication.
TR: The TR field 64 from the SARACC confirm indicates SUCCESS or
FAILURE in the transmission of a segment. TSEG: The TSEG field 60
in the SAR ACC request indicates whether the segment is the first
segment of the message (BOM), a continuing segment (COM), or the
final segment (EOM).
The states of the state machine diagram are as follows:
1.2 STATE S0:IDLE
State S0 is the initial state for all segmentation state machines.
In this state, no APP_SAR request is outstanding from the source
42.
S(00) Idle--When an SAR_ACC confirm is encountered in the idle
state for a previously confirmed message, the primitive is
discarded with no state change.
S(01) Sending--An APP_SAR request causes the initiation of the
machine and the transition to the Sending state. The segmentation
machine will remain in the Sending state until the message has been
sent and all expected responses have been received. A source
identifier is allocated.
1.3 STATE S1: SENDING
In state S1, the segmentation machine transmits segments of the
message until all segments have been transmitted.
S(10) Idle--The reception of a SAR_ACC confirm with TR=FAILURE
indicates a problem with the tramsission. Sending is aborted and
the machine, in transition to the idle state, discards outstanding
SAR_ACC confirms. The source 42 is notified with an confirm
(FAILURE) and the SI is deallocated.
S(11a) Sending--The segmentation machine forms segments from
message data, sets the appropriate T SEG and causes an SAR_ACC
request for each segment. The segmentation machine also checks the
CHECKSUM for the messgage. If the last segment is sent it
constructs the appropriately coded segment then sets flag
indicating last segment sent.
S(12) Wait Confirm--All segments have been sent and SAR_ACC
requests are outstanding. The segmentation machine waits for
confirms on any outstanding SAR_ACC requests.
1.4 STATE S2: WAIT CONFIRM
In this state, the segmentation machine waits for expected SAR_ACC
confirms.
S(20a) Idle--All SAR_ACC onfirms with TR=SUCCESS are received. An
APP_SAR confirm (SUCCESS) is generated indicating a best effort was
made of delivery. The SI is deallocated.
S(20b) Idle--SAR_ACC confirm with TR=FAILURE was received
indicating the attempt at delivery was unsuccessful. An APP_SAR
confirm (FAILURE) is generated indicating failure of delivery to
the source 42. The SI is deallocated.
S(22) Wait Confirm--An SAR_ACC confirm with SUCCESS is recorded in
the CONFIRMS_BACK array of flags.
The segmentation machine 50 described in the state diagram of FIG.
6 could be implementated by those skilled in the art using known
logic blocks or by programming of micro-computers.
An implementation of the reassembly machine 50 will now be
described with reference to the state machine diagram of FIG.
8.
2.1 ABBREVIATIONS USED IN THE REASSEMBLY MACHINE
RPT: Reassambly Packet Timer
SR RESOURCES AVAILABLE: Whether there are sufficient resources in
the reassembly machine to allow this reassembly machine to exit the
idle state.
CHECKSUM: Is a sum of all octects in a received message 20.
CHECKSUM GOOD: Indicates whether the checksum over message data
matches with the value for CHECKSUM count at the segmentation
machine 48 and transmitted to the re-assembly machine for error
checking.
LENGTH: The number of octects in the message field in the SAR_ACC
indication (BOM).
T_SEG: The T_SEG field in the SAR_ACC Indication indicates whether
the slot 32 is the first segment of the message (BOM), a continuing
segment (COM) or the final segment (EOM).
FIG. 9 diagrammatically illustrates the format for indication
primitives for communications between the access unit 4 and the
reassembly machine 50. In this case, the indication primitive 66
comprises a T_SEG field 68 and the slot 32. There is no need for
any control communication between the reassembly 50 and the access
unit 4.
2.2 STATE R0: IDLE
State R0 is the initial state for all reassembly machines. In this
state, no message is being reassembled.
R(00a) Idle--An SAR_ACC indication presents a segment with SSM set.
A single segment message is assembled. The checksum is computed and
if no error is indicated, an APP_SAR Indication presents the
complete message to the destination 46.
R(00b) Idle--An SAR_ACC Indication presents a segment with BOM set
and sufficient resources are not available to allow the reassembly
machine to leave the idle state.
R(01) Receiving--An SAR_ACC Indication presents a segment with BOM
set and sufficient SAR resources are available to allow the
reassembly machine to leave the idle state. Buffer allocated of
size in octects equal to LENGTH indication. Resources are allocated
and the data from the segment is buffered.
In FIG. 8, Action 1 beneath the transition line (11) includes the
steps of storing the received segment 32 in sequence and computing
the CHECKSUM over the message 20. Action 2 in transition lines
R0(00a) and R1(10b) includes of reassembling the message 20,
computing the CHECKSUM over the message 20. If CHECKSUM IN GOOD
then APP_SAR_IND, otherwise discard the message.
2.3 STATE R1: RECEIVING
In this state, the reassembly state machine is reassembling a
message. The machine will remain in this state until the complete
message has been assembled or until the Reassembly Packet Timer PRT
has expired or there is a problem with available resources. The RPT
ensures that the reassembly machine is not effectively put out of
service waiting for a lost EOM.
R(10a) Idle--When the Reassembly Packet Timer expires, the
reassambly machine will return to the idle state, discarding the
partially assembled message.
R(10b) Idle--When the complete message has been assembled (EOM
received) the reassembly machine computes the CHECKSUM. If the
CHECKSUM indicates no error, an APP_SAR indication presents the
received message to the destination 46. If the CHECKSUM indicates
an error, the reassambly machine returns to the Idle state and
releases resources.
R(11) Receiving--On each SAR_ACC indication with T_SEG=COM. the
reassembly machine will buffer the segment in sequence.
The reassemble state machine diagram shown in FIG. 8 could be
implemented in logic or by appropriate software.
FIG. 9 is another example of a state machine diagram for a
reassembly machine which is particularly designed for use with the
distributed queue system of the QPSX network disclosed in the
aforementioned international publication.
This receive machine has two states: IDLE and WAIT. In the IDLE
state the machine is not currently receiving any message. Thus in
this state the machine will check for slots 32 with TYPE field 36
equal to BOM or SSM. In the case a BOM code is received the machine
will check the DA in the information field. If the message is
addressed to the station the machine enters the second state. In
the case that SSM is detected in the IDLE state and the DA field
matches, the length and information fields are copied and the AU 4
will indicate the higher layers that a message is received. The
receive machine remains in the IDLE state after copying the slot.
The higher layers refer to the higher layers in the Open System
Interconnections (OSI) reference model.
The WAIT state is used to receive the slots 32 following the first
of a multisegment message. In this state the slots with TYPE field
36 equal to COM and the SI equal to that copied from the first slot
of the message will be received. The information fields of these
following slots are concatenated to form the complete message. New
messages addressed to the given station in this state are ignored
by the state machine. Further receive machines are required if such
messages are to be received. When the last slot of the message is
detected by the receive state machine, the machine will copy the
information field of the slot, indicating to the higher layers that
a message is received and return to the IDLE state. This completes
the receipt of the message. To guard against the loss of the EOM
slot, which would cause the receive state machine to be locked in
the WAIT state, the RPT timer is used. This timer is started after
each BOM or COM slot is received. If the timer expires before the
next slot is received, a failure in the transfer is assumed. The
machine will then clear all copied slots and return to the IDLE
state.
CONSTRAINED DESTINATION RESOURCES
In all practical implementations of destination facilities, i.e.
resources such as buffers and processing capacity will be limited
and loss of slots can occur unless a positive control mechanism is
introduced to control communication between source and
destination.
The method enables a source seeking use of the destination
resources for reassembly of the original message to be temporarily
held up until the required resources are available. The method
implements a variety of access disciplines. Among others, the
method supports first come first served queueing for the
destination resources described briefly below.
Control of access to destination resources is via a "ticket"
handout scheme in which the "ticket number" establishes the
position of source request in a distributed queue. The first
encapsulated segment of a message sent by a source is considered by
the destination as a request for resources. A "go ahead" message is
returned to the source if the required resources are available,
otherwise a "ticket" which uniquely identifies the sources position
in a distributed queue of requests is returned. When resources
become available the destination broadcasts the "ticket number" of
the next source to be serviced. Under the assumption that resources
are usually available, the delay incurred in waiting for the
"go-ahead" is avoided by the source continuing to send segments of
a message while it awaits the reply from the destination. In the
case of a negative reply, that is a "ticket" is received, the
source aborts transmission of further segments and resets its
transmission pointer to the first segment. Thus the source buffers
each message until it can be completely sent.
The TYPE field is used to control the generation of new ticket
number and thereby enables various resources access priority
schemes to be implemented. The TYPE field is also used to indicate
whether slots from part of a message whose transfer is
guaranteed.
FIG. 11 is a block diagram showing an implementation of a
reassembly machine 50. In this example, the machine is designed to
simultaneously handle seven messages and can therefore be regarded
as a seven-fold implementation of the reassembly machine which is
described in the state machine diagram of FIG. 8. The machine is
also able to simultaneously receive a single segment message.
The machine comprises a DA Select circuit 70 which receives fixed
length slots 32 from an access unit 4. The circuit includes an SI
comparator block 72 which also receives the slots 32 from the
access unit 4. Output from the comparator block 72 is connected to
the input of a buffer selector circuit 74. Outputs from the buffer
selector 74 are coupled to buffers 77 of a partitioned buffer space
76. The circuit also includes an SSM buffer 78 which receives
output from the DA select circuit 70. Outputs from the buffer 78
and buffers 77 are transferred to the destination 46 on output line
80, in the form of a reassembled variable length message 20.
The main function of the DA select circuit 70 is to check the
destination address within the BOM and SSM segments copied from the
slots 32 received by the access unit 4. If the destination address
matches that of the access unit then the DA select circuit 70 will
initiate the receipt and reassembly of the whole message provided
resource is available to do so.
The SI (Source Identifier) comparator block 72 comprises seven
identical sub-blocks (actually any number of sub-blocks may be
used. The number of sub-blocks specifies the number of multiple
segment messages 20 that may be reassembled simultaneously). The
function of an SI_Comp sub-block 82 is to copy all segments 40 with
the same SI code in the header of the slot 32. In this way all
segments 40 of a message 20 are received and allows the message to
be reassembled. When there are more messages destined to an access
unit than there are SI_Comp sub-blocks 82, the extra messages will
be lost.
The buffer selector 74 is a simple function that directs where a
copied segment should be stored. This block operates under the
control of the SI comparator circuit 72. Whenever any sub-block 82
detects that a segment should be received the buffer selector 74 is
directed to copy that segment into the associated buffer 77.
The buffers 77 are used to store segments 32 while messages are
being reassembled. The SSM buffer 78 is used to store all single
segment messages. Single segment messages do not need reassembly
since the complete message is contained within the one segment.
Hence the segment is copied directly to the SSM buffer 78. From
there it may be passed out as a reassembled message 20 to the
attached destination 46.
The seven buffers 77 of the buffer space 76 are each uniquely
associated with an SI_Comp sub-block 82 and each used to store all
segments 40 of a common message 20. Each buffer partition must be
of sufficient size so as to receive all segments of a message in
order to guarantee that no part of the message is lost. When all
segments of a message are received in a buffer partition then the
complete variable length message may be passed out of the buffer.
When the message is passed out, the SI_Comp sub-block 82 associated
with that buffer partition can be made free and hence used to
receive another message.
To further describe the operation of the reassembly machine it is
useful to consider the data flow paths through the machine when a
slot is passed to the machine from the network. A fixed length slot
32 copied from the access unit 4 will enter both the DA Select
circuit 70 and comparator block 72. The DA Select circuit 70 will
temporarily store the full slot 32. Each of the sub-blocks 82 will
only copy the SI field 38 of the slot.
The DA select circuit 70 will check the type field 36. If the type
is COM or EOM then the only action on the segment is to store it
temporarily, in case it needs to be transferred to the SI_Comp
block 72. If the type is BOM or SSM then the DA Select circuit 70
will check the DA address field 22 within the data segment 32. If
the DA address does not match that of the access unit 4, then the
segment is discarded. If the DA address does match that of the
access unit then the segment is intended for that unit. If the
segment is an SSM then the message part of it will be passed
directly to the single segment buffer 78. Hence the complete
message is received and no further action is required by the
reassembly machine except to pass it out as a reassembled
message.
If the segment received by the DA select circuit 70 is a BOM with a
DA match then the circuit 70 will check if there are any free
SI_Comp sub-blocks 82 available to control the receipt of the rest
of the segments of the message. If not the segment is discarded.
This event occurs when the reassembly machine is already receiving
messages on all sub-blocks. This extra message is hence lost.
If there is an SI_Comp sub-block 82 available then the DA select
circuit 70 will copy the SI from the segment and load it into the
free SI_Comp sub-block 82, hence making that sub-block busy. The
segment information field 40, that is the slot 32 minus the segment
header fields 34, 36 and 38 is then passed via line 84 to the
buffer selector circuit 70. The buffer selector ciruit 76 in turn
routes the segment 40 into the buffer 77 associated with the
claimed SI Comp sub-block 82. The association is signalled
explicity to the buffer selector circuit 74 from the SI_Comp block
82 via line 86.
For COM and EOM segments, that is the segments following the first
of a message, the decision to copy these into the buffer is made by
the SI_Comp circuit 72. Each sub-block 82 will compare the SI read
from the incoming slot via line 88 with its own SI. If no sub-block
has a match for the incoming SI then the segment is discarded. If
there is a match, the information segment 40 is passed from the
temporary storage in the DA select circuit 84 via line 84 to the
sub-block 82. As with a BOM segment the buffer selector circuit 74
will route the segment to the buffer partition associated with the
SI_Comp sub-block 82 that had a match for the SI. If the segment
passed to the buffer partition is an EOM segment the reassembly of
that message is complete. Hence the SI_Comp sub-block 82 is made
free and the complete message passed out along line 80.
FIGS. 12a and b are flow charts which illustrate the logical steps
involved in the reassemble machine 50, illustrated in FIG. 11.
Many modification will be apparent to those skilled in the art
without departing from the spirit and scope of the invention.
* * * * *