Docsity
Docsity

Prepare for your exams
Prepare for your exams

Study with the several resources on Docsity


Earn points to download
Earn points to download

Earn points by helping other students or get them with a premium plan


Guidelines and tips
Guidelines and tips

SA STUDY OF COMMUNICATION PROCESSOR SYSTEMS, Study notes of Immunology

S. APeMr'wl NOTES ... GTE Sylvania, Carnegie-Mellon University, Honeywell, and RCA Corporation. ... CHAPTER III CARNEGIE-MELLON STUDY OF MULTI-PROCESSOR.

Typology: Study notes

2022/2023

Uploaded on 05/11/2023

larryp
larryp 🇺🇸

4.8

(34)

353 documents

1 / 196

Toggle sidebar

This page cannot be seen from the preview

Don't miss anything!

bg1
RADC-TR-79-310
Final
Technical
Report
December
1979
-
SA
STUDY
OF
COMMUNICATION
PROCESSOR
SYSTEMS
I
Wayne
State
University
Tse-yun
Feng
nharma
P.
Agrawal
[APPROVED
FOR
PUBLIC
RELEASE;
DISTRIBUTION
UNLIMIYED]
DTiC
i'
'~
FF6
L
b
96
ROME
AIR
DEVELOPMENT
CENTER
Air
Force
Systems
Command
Griffiss
Air
Force
Base,
New
York
13441
80
2
25
026
pf3
pf4
pf5
pf8
pf9
pfa
pfd
pfe
pff
pf12
pf13
pf14
pf15
pf16
pf17
pf18
pf19
pf1a
pf1b
pf1c
pf1d
pf1e
pf1f
pf20
pf21
pf22
pf23
pf24
pf25
pf26
pf27
pf28
pf29
pf2a
pf2b
pf2c
pf2d
pf2e
pf2f
pf30
pf31
pf32
pf33
pf34
pf35
pf36
pf38
pf39
pf3a
pf3b
pf3c
pf3d
pf3e
pf3f
pf40
pf41
pf42
pf43
pf44
pf45
pf46
pf47
pf48
pf49
pf4a
pf4b
pf4c
pf4d
pf4e
pf4f
pf50
pf52
pf53
pf54
pf55
pf56
pf57
pf58
pf59
pf5a
pf5b
pf5c
pf5d
pf5e
pf5f
pf60
pf61
pf64

Partial preview of the text

Download SA STUDY OF COMMUNICATION PROCESSOR SYSTEMS and more Study notes Immunology in PDF only on Docsity!

RADC-TR-79-

Final Technical^ Report

December 1979

-

SA STUDY^ OF^ COMMUNICATION PROCESSOR SYSTEMS

I Wayne State University

Tse-yun Feng

nharma P.^ Agrawal

[APPROVED FOR^ PUBLIC^ RELEASE;^ DISTRIBUTION^ UNLIMIYED]

DTiC

i' '~ FF6 L b 96

ROME AIR^ DEVELOPMENT^ CENTER Air Force^ Systems^ Command Griffiss Air^ Force^ Base,^ New^ York^13441

80 2 25 026

This report has been reviewed by the RADC Public Affairs Office (PA) and Is releasable to the National Technical information Service (NrIS). At NTIS it will be releasable to the general public, including foreign nations.

RADC-TR-79-310 has been reviewed and is approved for publication.

APPROVED:

t SES L. PREVITE Project Engineer I

APPROVED: ' K .Cju

WENDALL C. BAUMAN, Colonel, USAF

Chief, Information Sciences Division o

FOR THlE COMANDER:c

'-JOHN P. HIUSS

Acting Chief, Plans Office I

If your address has changed or if you wish to be removed from the RADC

mailing list, or if the addressee is no longer employed by your organization,

please notify RADC (ISCA) Griffiss AFB NY 13441. This will assist us in

maintaining a current mailing list.

Do not return this copy. Retain or destroy.

-Mi I

UNCLASSIFIED SECURITY CLASSIFICATION OF THIS PAOC1wen Data Entered)

fis given to those^ systems^ which^ have^ recently^ been proposed^ by^ various organizations to^ the^ government^ agencies.^ Such^ a^ restriction^ is^ justified^ by the -fact that a new evolved architecture^ is^ expected^ to^ overcome^ most^ of^ the shortcomings and^ limitations^ of^ the^ existing^ systems.^ General^ descriptions^ of the systems included here are^ those^ proposed^ by:^ North^ Electric^ Company, GTE Sylvania, Carnegie-Mellon University, Honeywell, and RCA Corporation. An attempt is made to obtain a comparative statement^ of^ various^ parameters^ and the uniqueness cf these systems so that their^ salient^ features^ and^ deficiencies can be compared. Finally, our recommendations and architectural strategies Vare outlined to provide guidelines for^ CPS^ design.^ Further^ thorough investigation is necessary in^ order^ to^ obtain more^ precise^ architectural requirements of CPS's.

Accession~ Fcr -_ NTIS GP,: l 1 D TAB. \nnumouncedj JustificatioX..

DistribuL.^ ..

Di ,

UNCLASSIFIED SECURITY CLASSIFICATION OF^ THIS^ PAGE1"Whon^ Data^ Entered)

TAB3LE OF CONTENTS

LIST OF FIGURES. .i ........ i

LIST OF TABLES..... .. ..................... V

7CHAPTER I^ GENERAL^ CHARACTERISTICS

1.1 Introduction................. ..... .. .. .. ... .. 1

1.2 Communication Considerations. ......^ ..........^2

1.3 Architectural Consdierations. ..... .......... 6

1.4 Report Organization. .. .......... ......... 9

CHAPTER II NETW4ORK ANALYSIS CORPORATION STUDY OF INTEGRATED

DOD VOICE AND DATE NETWORKS

2.1 Introduction .. ..... ............. ...... 10

2.2 Description of Switching Strategies.... .. . 1

2.3 Cost Comparison of Switching Technologies ........ 15

2.4 Discussion .. .... ........... ......... 21

CHAPTER III CARNEGIE-MELLON STUDY OF MULTI-PROCESSOR

ARCHITECTURE

3.1 Introduction .. ..... ........... ....... 22

3.2 Performance Evaluation .. ........... 22

3.3 Reliability and Life Cycle Cost Modeling .. *.

3.4 Conclusion ..... .,.... , 43

CHAPTER IV NORTH ELECTRIC COMMUJNICATIONS PROCESSOR SYSTEM

4. ntroduction...^.^.^.^.^.^^4

4.2 Architectural Description. .. ,.. 4.3 Performance Characteristics t^ t^ q^ f^^64

4.4 Discussion.... ........ 63

CHAPTER V GTE SYLVANIA SENET-DAX STUDY

5.1 Introdu inion................,.,.

5.2 Architectural Description. *

5.3 Performance (^) Characteristics....... , 89

CHAPTER VI CARNEGIE-MELLON C.mmp, AND Cm*

6.1 Introduction ......^ ,^9 6.2 Description. 99 6.3 Performance Characteristics ..... 108

6*4 iscuoion... ... 11

Muliplxed LIST OF FIGURS Pg

2.1 Examplary Hybrid Switching Channel Structure: Time

DivsinMltpleedFrame. .. .. . ..... ......... 12

22Integrated versus Segregated Voice and Data Networks

as a Function of Voice Digitation Rate and SWitching

Technology. .. ... . ....... .. ... ....... .... 14

2.3 -Monthly Backbone Switching and Transmission Cost as i

__ a Function of Voice -Digitation Rate and Switching

Technology (^) .. .. .. .. .. .. .. .. .. .. .. (^) .... 16

2.4 Examplary Packet Voicze/Data Network and Information

Flow. .. ... ...... ..... . .... ......... 19I

3.1 An Abstract View of a Multiproces-sor. .. ... ......... 23

3.2 Carnegie Mellon C.mmp....... ....... ..... .. .. ... .... 2

3.3 Pluribus System .. .... ... ...... .......... 4

3.4 Carnegie Mellon Cm* .. .... .... ...... . ....... 5

3.5 'GlobalBus System.. ... ..... ....... ....... 2 6

3.6 Tandem System.... ...... ..... ..... ..... 26

3.7 Process Identification .. .. .. ...... ..... ...... 2 7

3.8 C.mmp. with 64 Kw per port, 512 Kw required (lumped

switch)......................... 33

3.9 C.mmp. with 64 Kw per port, 12 processors required

(distributesi ch... .. .. .. .. ..... . ..... 34

3.10 Pluribus with 6 processor buses, 3 memory buses, 128 Kw/

memory bus, 216Kw required.... .. .. ............ 3

3.11 Pluribus with 6 p.rocessor buses, 3 memory buses, 128 Kw/

__ memory bus, 8 processors required ........... 36

3.2M -Clse Cm* with thre posors/cluster, '32 Kw/Cm,

__144 (^) Kw required (parallel buses).. .. .. .... (^).. 37

3.13 3 Cluster (m* with three processor) cluster, 32 Kw/Cm,

__144 1 r w required (checkboard) ...............^38

__3.14 Global Bus with 12 processors, 32 Kw/processor, 128

__Kw required .....................

NO- r-~

3.15 Global Bus with 12 processors. 32 Ky/processor, (^) B processors required ........ ... (^) ...... 40 3.16 Tandem with 12 processors, (^) 32 Kw/processor, 16 Ky/processor

required .......... ............ 41

i 3.17 Tandem with 12 processors, 32 Kwprocessor, 8 processors required q ........ ..^.^ ........^^42 3.18 Equivalent number^ of^ processors^ and^ memory^ modules^.^.^ .- U 3.19 Equivalent performance 62% voice, (^241) data, normal processor. 24" Ti carrier in each line, three lines......****.****... ...^5 3.20 Equivalent (^) performance 624 voice, 28 data ten ties faster processor, (^) 48% of Ti carrier in each line, three lines .4.. 46 3.21 Equivalent (^) performance 62% voice, 24% data ten times faster

processor, 24% of TI carrier in each line, three lines .... 47

IV


==Figures (^) Page

8.&6 Three Node Subnetwork Function Configuration 154

8.7- Node Simulator Configur-cion. .. ..... ..... ...... 156

9.1 ESS Bus Confuguration....... ... .... ......... 165g

9.2 STS-4 Basic Logical Structure,......... . ...... 168

vi

LIST OF TABLES

Table Page

1.1 Subscriber Classes and Associated Parameters 4 2.1 Backbone Network Costs............. .. 17 R. 2.2 Unit of Cost of Alternative Network Technologies ... .... 18 2.3 Cost Below Which it is Economical to Provide VTDR Devices in the Backbone Network ..... ..... ..... .. 20 I 2.4 Sample Results Comparing Total System Monthly Cost for the Option of Digitization in the Handset for 1000, Handsets ..... ...... ...........................^^20 3.1 Amount of Traffic in each Communication Line ... ......... 29 3.2 Memory References for Various Processors .. ........... .. 29 3.3 Minimum number of Processors needed for 62% voice and 28: data, Decomposition I ........ ..................... 30 3.4 Minimum requirements for 62% voice, 28% data, Decomposition II 30 3.5 Minimum requirements for 25% voice, 67- data, Decomposition 1 30 3.6 Minimum requirements for 25% voice, 67% data, Decomposition 11 31 3.7 Minimum number of processors required for 62% voice, 28% data with Decomposition I and ten times faster memory and Processor 31 3.8 Minimum requirements for 62% voice, 28% data, Decomposition II tentimes faster Processor and memory ... ....... .... .. 31 3.9 Minimum requirements for 25% voice, 67% data, Decomposition I, ten times^ faster^ Processor^ and^ memory^ ...^ ...........^ ..^32 3.10 Minimum requirements for 25% voice, 67% data, Decomposition II, ten^ times^ faster^ Processor^ and^ memory. ....^ ...........^^32 4.1 Typical Examples of Parameters for Master Model .... ....... 65 -- 4.2 Traffic Characteristics of Systems ...... .............. 66

4.3 Circuit Switched System. ... .....^ ........^ ....^^67 4.4 Syat~m Features .... ....... .................. .... 67 5.1 Maintenance M ethodology ...... .................... ..^79 5.2 SENET-DAX Subscriber Service Features ....^ .............^ ...^83

vii

_ = -- = -M=

EVALUATION

Many Coitmiunication Processors (CP) were not designed specifically

for c:mmunication processing but were selected from commerically

available processors and adapted to this application. Performance

deficiencies relating to these processors are therefore to be

-expected. These deficiencies along with future communication

demands require careful evaluation in light of today's dramatically

changing technology so that cost-effective Communication Processor

Architecture designs may be effected.

The recent advances in LSI technology has caused significant

changes in the field of computer architecture. One trend is to

construct a processing system by interconnecting a large number

of processors and memory modules. Key to this approach is the design

of an efficient intercommunication network which can avoid the bottle-

necks introduced by the traditional b systems and enhance its

system performance in terms of availability, reliability, and

throughput. Also involved is the processing unit design. The

processing unit should include features for enhancing processing

power within chips, improving testability and system reliability.

__ Similar considerations can be used to design the CPS. However, A."

the performance of a CPS is only partially dependent on the computa-

tion effectiveness, thus architectural decisions have to be taken

to fulfill the communication requirements.

ix

_C (^) --- (^) _A-

This report discusses the commurication and the architectural

considerations that characterize the salient features of a CPS.

In addition, it provides the relative advantages of different

approaches currentlypprevalent in existing or proposed systems.

This effort's significance is both to architectural trends

which take advantage of low cost commercially available microproces-

sor technology and to the application of this technology to the field

of communications. -As such, it has relevance to TPO's R3A and R3D,

This report will be disseminated to the comunications

community. Follow-on efforts will focus strictly on the multistage

interconnectiw strategy for architectures in general.

J ES L. PREVITE

Project Engineer

X

I - - ~-- ~ -~ -- T

effectiveness of^ the^ communication^ handling^ techniques.^ The^ design

objective of the computing system is only partially dependent on the

computation effectiveness and the architectural decisions have^ to^ be^ II

taken to fulfill the communication requirements [2]. This leads to

an approach which differs from thea conventional design criterion.

In the following, we discuss the communication and the architec-

tural considerations that characterize the salient features of a CPS.

In addition* we provide the relative advantages of different approaches

currently prevalent in the existing or in the proposed systems.

2. Communication Considerations

Several factors such as the communication technique, the commu-

nication network characteristics the network structure, and the network

S- design need to be examined when designing (^) a communication system. =

Those factors are briefly^ discussed^ below.-

A. Communication Technique

The communication techniques are critically important as

there could be more than one decision point along a path set up

i an intercommunication networkt There are at least two commu-

nication levels:^ source-to-destination^ and^ switching-point-to-^ A

switching-point. There are three possible switching methodologies M

for supporting general networking requirements:

1. circuit switching,

2. packet switching, and

3. integrated^ circuit-packet^ switching.

The circuit switching involves the establishment of a dedi-

dated path between any two terminals and the connection is dis-

rupted only when ony one of the two users signals for such action.

This switching technique is relatively inefficient for transmission

o-of short messages. But, it is more suitable for transfer of

bulk data.

In contrast to circuit switching, packet switching attempts

i to multiplex the usu of the communication circuit among all

related terminal unts. Messages are typically broken into a --N

series of fixed length, addressed packets of data which are

routed independently to their destination (^) using store-and-forward procedures. The packet switching cain partially solve the blocking problem of interconnection networks. However, it also increases the complexity of the control procedure and the effective time I

delay.

The integration of circuit and packet switching into a unified switching concept [3] enables efficient and economical use of the communication system. It also allows (^) a better connec- tion between the terminals. This Slotted Envelope NETwork (SENET) concept utilizes constant perio', self-synchronizing master frame to convey messages to the multi-node network. The assignment of compartments to different types of traffic provides both circuit switching and packet switching capabilities simultaneously. The characteristics of different subscribers have been used to separate three different classes of traffic [3]. These classes and associated parameters are given in (^) Table 1.1. The size of the compartments for each group of traffic is allowed to vary, dynamically with the instantaneous traffic conditions and is achieved by performing the frame composition and decomposition at each node. Each envelope begins with a header and the end

is recognized by a trailer.

B. Communication Network Characteristics The characteristics which can be used to specify the inter- communication network are identified as follows:

1. topology of interconnnction network,

  1. control structure of interconnection network,

3. logical complexity,

4. blocking probability,

  1. message response^ time,

6. system capacity or throughput,

-- 7. network reliability,

  1. sensitivity,
  2. traffic bottleneck or deadlock, 10. transmission error rate, and 11. cost.

3 1

t (^) ________

I C. Network Structure

The cast and performance of the intercommunication network

depends greatly on the network structure as the designer (^) chooses.

The network (^) structure in the form of cross-bar switches is a widely utiliied circuit-switching technique for establishing a connection between any two users. The maximum number of^ simul-^ V taneous connections set-up between various subscribers can be

achieved only if each cross-point is capable of providing conflict-

free parallel switching. Since the number of cross-points grows exponentially (^) with the number of subscribers, the cost of the circuitry required for the switching facilities becomes signifi- cantly high when the number of (^) subscribers is large. The solution

to this problem at the cost of set-up time is the use of multi-

stage network structures where the network is partitioned into

several stages (^) and the desired connection is established via intermediate stages. The cost advantage-offered (^) by the multi- stage network is partially off-set (^) by the control selection time. However, multistage network usually provide multiple path connec- tions and are, therefore, more fault-tolerant than (^) cross-bar

switches. The current trend is to utilize the multistage network

as effectively as possible.

There are four types of multistage interconnection networks

1. strictly nonblocking network,

2. wide-sense nonblocking network,

3. rearrangeable (^) nonblocking network, and

  1. blocking network.

A strictly nonblocking network allows connection between

any two lines regardless of the current connections. A wide-

sense nonblocking network provides the same service if specific

routing rules are observed. The rearrangeable nonblocking network

also (^) has the same capability except that it requires alterations in the existing connections, The blocking network can perform

many but not all possible connections.

Generally speaking, the nonblocking property is not the prime

criterion for choosing network structure. Network -tructure with

_ 5 [

properties of easy manufacturing, low cost, simple control, = graceful degradation, and short delay time are the major parameters.

D. Network Design The hardware design considerations for the interconnection network are:

1. cross-bar switches versus multistage design,

  1. central versus distributed control, 3. linkage versus connection capability,
  2. circuit partitioning and chip types,
  3. number of interconnectioi points versus system relia- bility, and

6. synchronization.

The network software design is essentially to have a set of basic (^) control procedures to ensure an efficient, correct and smooth transfer of information in the interconnection network system. The basic control procedure can be partitioned into the following four categories [5,61: I. communication protocols,

  1. flow control procedures,
  2. graceful degradation^ considerations,^ and
  3. routing procedures.
  4. Architectural Considerations The evolution of the CPS architecture shows that the trend has been from uniprocessor architecture to multiple-processor architecture i either in circuit switching [7-12] or in stere-and-forward switching [13-14] and the interconnection technique tends to be full connection [15]. Current proposals [12,16-21] of CPS architectures also show that a CPS with the capability of performing integrated circuit and

store-and-forward sw34tching (e.g. SENET) is preferred. Thus, various attributes of interconnection organizations such as the transfer strategy, the control method, the path structure, and the system topology are used to classify the actual system designs [1]. of a Basedcomputer.^ on^ thethe^ interconnections present day multiple-processor^ between^ various systemsfunctional can^ beunits

i °