Chapter 4

Electronic Space Division

Switching

School of Information Science and

Engineering, Shandong University
Associate Prof., Deqiang Wang

Outline

Stored Program Control

Centralized SPC
Distributed SPC
Software Architecture
Application Software
Enhanced Services
Two-Stage Networks
Three-Stage Networks
N-Stage Networks

Introduction
Efforts to improve the speed of control
and signaling
Late 1940s and early 1950s. Use of
vacuum tubes, transistors, gas diodes,
magnetic drums and cathode ray tubes.
Arrival of modern electronic digital
computers.
The registers and translators of common
control systems can be replaced by a
single digital computer.

4.1 Stored Program Control

Stored Program Control (SPC)
Carrying out the exchange control
functions through programs stored
in the memory of a computer.
Consequence
Full-scale automation of exchange
functions
Introduction of a variety of new
services

4.1 Stored Program Control

New features possible for SPC

Common Channel Signaling (CCS)

Centralized maintenance
Automatic fault diagnosis
Interactive human-machine interface

Special Requirements of SPC

Operating without interruption
Fault tolerant hardware and software

4.1 Stored Program Control

Two types of SPC switching system
Electromechanical Switching
SPC+Electromechanical switching network
Electronic Switching
SPC+Electronic switching network

4.1 Stored Program Control

Organization of SPC
Centralized SPC
Broadly used in early SPC switching systems.

Distributed SPC
Gaining popularity in modern switching
systems.

4.2 Centralized SPC

Concept
All the control equipment is
replaced by a single powerful
processor.

Configuration of centralized SPC

Typical organization
Redundant configuration

4.2 Centralized SPC

Operation modes in redundant
configuration (e.g. dual processor)
Standby () mode
Synchronous duplex () mode
Load sharing () mode

Standby mode
How does it work?
All processors have the same capability
to control the switching procedure.
One processor is active and the other is
on standby, both hardware and software
wise.
The standby processor is brought online
only when the active processor fails.

Standby mode
How does the standby processor take
over the control properly?
State of the exchange system should be
clear to the standby processor as its
starting point.
Which of the subscribers are busy or
free?
Which of the trunks are busy or free?
Which of the paths are connected
through the switching network?
So on

Standby mode
Reconstitution of the state
Scanning:
The standby processor scans all status
signals as soon as it is brought into
operation.
Only the calls which are being established
at the time of failure are disturbed.
Only feasible for small exchanges.

Shared secondary storage: popular.

Standby mode
Shared secondary storage:
The active processor copies system status into a
secondary storage periodically, say very 5
seconds.
As soon as a switchover occurs, the online
processor loads the most recent update of the
system status from the secondary storage and
continues the operations.
Only the calls which changed status between the
last update and the failure are disturbed.
Feasible for large exchanges.

Synchronous duplex mode

How does it work?

Both two processor execute the same set of

instructions.
One of the processor actually controls the
exchange.
The results from two processors are compared
continuously by a comparator.
If the results match, the system works normally.
Otherwise, a fault occurs, a check-out program
is run independently in both two processors to
determine which one is faulty.
The faulty processor is taken out of service, and
the other one works independently.

Synchronous duplex mode

In case of transient failure of
comparator, there are three
possibilities exist:
Continue with both processors.
Take out the active processor and continue
with other processor.
Continue with the active processor but
remove the other processor from service.

Load sharing mode

How does it work?

Both two processors have access to entire

exchange environment. Each of them has
independent memories for redundancy purpose.
Both two processors are active simultaneously
and share the load and the resources
dynamically.
An incoming call is assigned randomly or in a
predefined order to one of the processors which
then handles the call right through completion.
Inter-processor links are configured for
processors to exchange information needed for
mutual coordination and verifying the state of
health of the other.

Load sharing mode

If a processor fails, the other
processor takes over the entire load
including the calls already set up by
the failing processor.
Exclusion mechanism in resource
sharing
The processors should not seek the same
resource at the same time.
Implementation: hardware & software

Load sharing mode

Traffic distribution between processors
Load sharing increases the effective
traffic capacity by 30 percent
compared with synchronous duplex.

Availability of system
A telephone exchange must show
more or less a continuous availability
over a period of 30 or 40 years.
Availability ()
Availability of a single processor system
A=MTBF/(MTBF+MTTR)
MTBF: mean time between failures
MTTR: mean time to repair, MTTR<<MTBF
normal

normal

normal

fault

normal

fault

normal

Availability of system
Unavailability of a single processor system
U=1-A=MTTR/(MTBF+MTTR)MTTR/MTBF

Availability of dual processor system

Mean time between failures
MTBFD=MTBF2/(2*MTTR)
AD=MTBFD/(MTBFD+MTTR)
=MTBF2/(MTBF2+2MTTR2)
UD=1-AD 2MTTR2/MTBF2
normal
normal

normal
fault

normal

normal

fault

normal

normal

fault

normal

normal

fault

normal

Availability of system
Example:
MTBF=2000hours, MTTR=4hours
U=? UD=?
Solution:
U4/2000=2x10-3
i.e. 525 hours in 30 years.
UD2x16/(2000x2000)=8x10-6
i.e. 2.1 hours in 30 years.

Coordination of functions
Priority of Control functions
Level 1: O&M and charging
Level 2: Call processing
Level 3: Event monitoring and distribution

Interrupt processing

Vectored interrupt ()
The interrupting source supplies branch address
information to the processor.
Nonvectored interrupt
A main interrupt service routine scans the
interrupt signals and decides on the appropriate
routine to service the specific interrupt.

Levels of control functions

Interrupt processing

4.3 Distributed SPC

Concept of distributed SPC
The control functions are shared by
many processors within the exchanges.

Background
Low cost processors

Advantages
Better Availability ()
Better Reliability

4.3 Distributed SPC

Decomposition of Control Functions
Vertical decomposition ()
The exchange environment is divided into
several blocks.
Each block is assigned to a processor.
A processor performs all control functions
related to the corresponding block.
The processor in each block may be duplicated
for redundancy purposes.
Obviously, the control system consists of a
number of control units.
The modular structure is flexible for system
expanding.

4.3 Distributed SPC

Exchange environment
Block

Block

Block

Unit 1

Unit 2

Unit n

Control System

Vertical decomposition

4.3 Distributed SPC

Horizontal decomposition ()
The control functions are divided into
groups, e.g. event monitoring, call
processing, and O&M functions.
Each processor performs only one or some
of the exchange control functions.
A chain of processors are used to perform
the entire control of the exchange.
The entire chain may be duplicated to
provide redundancy.

4.3 Distributed SPC

4.3.1 Level 3 Processing

Tasks
Scanning: Subscriber lines/Trunks
Distribution: Subscriber lines/Trunks
Marking: Registers

Location
Close to switching network, junctors and
signalling equipment.

4.3.1 Level 3 Processing

Control schemes
Hard-wired control unit
Designed as a collection of logic circuits
using logic elements

Microprogrammed control unit

Using microprocessors to implement all
control logic

4.3.2 Level 2 Processing

Task
Call processing
Level 2 processor = Switching processor

Type of processors used

House-developed ()
Standard

4.3.2 Level 2 Processing

Specific characteristics of Switching
processors
Efficiently designed instructions
Special instructions
Special architecture for availability, fault
tolerance and security
Special peripherals for telephones
Special communication link for remote
control

4.3.2 Level 2 Processing

Traffic handling capacity
t=a+bN
N=(t-a)/b
t: the load of the switching processor
measured by its occupancy.
a: fixed overhead () depending upon
the exchange capacity and configuration.
b: average time to process one call.
N: number of calls per unit time.

4.3.3 Level 1 Processing

Q&M functions involved
Administer the exchange hardware and software.
Add, modify or delete information in translation
tables.
Change subscriber class of services.
Put a new line or trunk into operation.
Monitor traffic.
Detect and locate faults and errors.
Run diagnostic and test programs.
Man-machine interaction.

4.4 Software Architecture

Software of SPC systems
System software
Application software

We focus on the call processing

software systems.

Operating system of Call processing

Multiprogramming () Feature
Call processing is event oriented.
Call setup involves a series of separated
processing actions, each action is
triggered by an event occurring at a
subscriber line or trunk.
There are many calls being processed
simultaneously, each handled by a
process.
Process: a program in execution.

Process states and transitions

States of a process

Running state: currently possessing the CPU

resources.
A CPU can be allocated to only one process at
any instant of time.
Ready state: waiting for availability of the CPU.
All processes in ready state are ordered
according to some priority.
The first process in the ordered queue is the
next which will receive the CPU.
A timer is employed to prevent any process
from monopolising the CPU.
Blocked state: waiting for some event to occur.

States and transitions

Representation of a process
Process Control Block (PCB)
PCB is a data structure containing relevant
information of a process
Current state of the process
Process priority and CPU scheduling parameters
Register save area to save the contents of the
CPU registers when an interrupt occurs
Memory allocated to the process
Process account like the CPU time limits and
usage, process number etc.
Status of events and I/O resources associated
with the process.

Process switching ()
Process switching occurs in following
cases:
The current running process becomes
blocked;
An event or interrupt triggers a high
priority process.

Process switching implies PCB saving

and loading.

Mutual exclusion of processes in

accessing shared data
Data sharing in switching system
Processes share common variables,
update common tables, write into
common files, and so on.
Mutual exclusion: When one
process accesses a shared table, all
other processes want to access the same
table are kept waiting.

Software production
Basic factors associated with switching
software

Complexity and size of the software

Long working life required
Real time operation
Stringent reliability and availability
Software portability

Four stages in software production

Functional specification
Formal description and detailed
specification
Coding and verification
Testing and debugging

CCITT standards on language

Z.100
specification description language (SDL)

Z.200
CCITT high level language (CHILL)

Z.300
Man-machine language (MML)

4.5 Application software

Classes of application software
Call processing software
(core resident: )

Administrative software
Maintenance software

Description of a software package

Organization
Data structure
Processing functions

Organization of switching software

Modular organization()
The software package is divided into program
modules.
Each module deals with a specific task.
The size of each module varies depending on the
task.
Modules exchange data through interfaces or
data table.
Modules are grouped together to constitute
independent functional units.
A functional unit runs as a separate process.

Organization of switching software

How to string up modules within a process?
Special program
Chaining table
Each entry consists of a key and a module
number
When a module completes execution, entries in
the chaining table are scanned and the keys are
compared with a function status key.
The matched module is executed next.
Flexibility for modifying a functional unit:
adding or removing modules.

Data structure of switching software

Parameters
System parameters
Afford flexibility at the overall system level,
such as signalling time delays, fault
thresholds, etc.

Office parameters
Define program execution limits at specific
exchanges, such as the number of
subscribers, the maximum number of
simultaneous calls etc.

Data structure of switching software

Data type and storage
Semipermanent data:
e.g. hardware parameters
Updated rather infrequently
Stored as files

Temporary data
e.g. information specific to a process
Have a lifetime equal to the process they
pertain to
Stored as tables

Data structure of switching software

Hardware description files
Terminal circuit connection file

Associated distributor address

Associated scanner address
Switching network address
Physical location address

Switching network configuration file

Connected subscriber lines
Trunks towards other exchanges
Rules for digit translation and routing

Data structure of switching software

Subscriber line information files
Correspondence between line equipment
number and directory number
In-service or out-service status
Type of telephone instrument
Nominal transmission parameters
Call restrictions
Subscribers entitlement
Subscriber instructions

Data structure of switching software

State description tables

Subscriber line state table

Terminal circuit state table
Switching network link state table
Working area of a process

Functions of switching software

Call processing

Line scanning
Calling line related functions
Routing and link setup
Called line related functions

Administration
Traffic monitoring/control/billing

Maintenance
Diagnostic / preventive maintenances

4.6 Enhanced Services

Categories of enhanced services
1. Services associated with the calling
subscriber and designed to reduce the
time spent on dialing and the number of
dialing errors.
2. Services associated with the called
subscriber and designed to increase the
call completion rate.
3. Services involving more than two parties.
4. Miscellaneous services.

Category 1
Abbreviated dialing ()
Recorded number calls or no dialing
calls ()
Call back when free ()

Category 2
Call forwarding ()
Operator answer

Category 3

Calling number record ()

Call waiting
Consultation hold
Conference calls

Category 4
Automatic alarm
STD barring
Malicious call tracing
STDsubscriber trunk dialing

4.7 Two-Stage Networks

Theorem
For any single stage network, there
exists an equivalent multistage network.

Simple Two-stage NxN network

A NxN single stage network with a
switching capacity of K connections can
be realized by a two-stage network of
NxK and KxN.

Simple Two-stage NxN network

First Stage: Any of the N inlets can be connected to any of the K

outputs. NK switching elements.
Second Stage: Any of the K inputs can be connected to any of
the N outlets. NK switching elements.
There are K alternative paths for any inlet/outlet pair connection.

Simple Two-stage NxN network

Full connectivity/full availability
Any of the N inlets can be connected to
any of the N outlets.

Example
Assume 10% of the subscribers to be
active on average. Set K to be N/16. The
number of switching elements is S=N2/8.
For N=1024, we have K=64, S=131072.

Note: Feasibility & Flexibility

4.7 Two-Stage Networks

Single stage vs. Multistage networks

Inlet to outlet connection

Quality of link
Utility of cross-points
Establishment of a specific connection
Cross-point & path Redundancy
Number of cross-points
Capacitive loading problem
Blocking feature
Call establishing time

General two-stage networks

Terminology
Expanding network: M<N
Concentrating network: M>N
Square network: M=N

M inlets

N outlets

General two-stage networks

Architecture of General two-stage networks
Multiple small size matrices are used in each
stage.
Easy to be realized in practice.
Flexible in system design.

MxN two-stage network design

Decomposition: M = p x r, N = q x s
Switching matrices: p x s and r x q
Full availability: There must be at least one out
let from each block in the first stage terminating
as inlet on every block of the second stage.

First stage

Second stage
Switching Matrices

General two-stage networks

Parameters
Number of switching elements
S=psr+qrs=Ms+Nr

Switching Capacity
i.e., the number of links between the first
and the second stages.
SC=sr

General two-stage networks

Parameters
Blocking probability
Blocking condition 1
9 There are rxs calls in progress, and the (rs+1)-th
call arrives;
9 The blocking probability PB is dependent on the
traffic statistics.

Blocking condition 2
9 There is a call in progress from I-th block in the first
stage to the J-th block in the second stage, and
another call originating in the I-th block destined to
the J-th block.
M ( s 1)(( M / r ) 1)
9 Blocking probability
P =
B

rs ( s 1)

General two-stage networks

How to choose values of r and s?
Both S and SC are proportional to r & s.
Blocking probability PB is reversely
proportional to r & s.
Strategy: Tradeoffs should be made
between cost and quality of service.
The values of r & s should be as small as
possible but give sufficient links to provide
a reasonable grade of service to subscribers.

Square two-stage networks

Baseline networks
Square switching matrices are used as building
blocks.
p=r=s=q=N1/2
There are N1/2 blocks, each block is a switching
matrix of N1/2xN1/2 inlets and outlets.
Switching elements: S=2NxN1/2
Switching capacity: SC=N
Support N simultaneous calls only if the traffic is
uniformly distributed.

Baseline networks

N
inlets

N
outlets

Square two-stage networks

Nonblocking networks

Why does blocking occur?

Only one link exists between a pair of first stage
and second stage blocks.
How to reduce the probability of blocking?
Provide more links between the first stage and
second stage blocks.
How many links should be provided?
A group of k=N1/2 links should be provided for
each pair of first stage and second stage blocks.
S=2N2.
In comparison with single stage network, the
number of switching elements is doubled.

Nonblocking networks

N
inlets

11

11

21

21

1
N

1
N

S=kx2NxN1/2=2N2

N
outlets

Remarks on two-stage networks

A standard way of designing blocking
networks with full availability.
The expanding and concentrating network
structures of small size can be implemented
easily.
Alternative paths for establishing a
connection can be provided by two-stage
networks, but with more switching
elements in nonblocking design.

Question
Is there any way to implement
blocking & nonblocking networks with
even less switching elements?
Solution: three-stage or multi-stage ~

4.8 Three-Stage networks

General structure of an NxN threestage blocking network

Stage 1: p x s switching matrices

Stage 2: r x r switching matrices
Stage 3: s x p switching matrices
N=pxr, s is changable
Compared with a two-stage network,
there are s alternative paths between a
pair of inlet and outlet.

NxN three-stage blocking network

Number of switching elements
S=rps+sr2+spr=2Ns+sr2=s(2N+r2)
If square matrices are used in both the first and
third stages, then p=s=N/r and S=2N2/r+Nr.
For a given value of N, there exists an optimal
value of r which minimizes the value of S.
The optimal value of r is r=(2N)1/2 and the
corresponding minimum of S is Smin=2N (2N)1/2
p=N/r=(N/2)1/2

How about the switching capacity?

NxN three-stage blocking network

Blocking probability
analysis

Probability graph
Circle: stage
Line: link
A graph can be broken
down into serial and
parallel paths
Notation
: probability that a link is
busy.
=1-: probability that a
link is free.

Blocking probability analysis

When s serial links complete a
connection, the blocking probability is
PB=1-(1-)s
If there are s parallel links, the
blocking probability is
PB=s

Blocking probability analysis

The blocking probability of NxN three-stage
network
If denotes the probability that a inlet at the
first stage is busy, then =p/s = /k, where k=s/p
There are s parallel paths
What is
Each path consists of two serial links

Blocking probability of each path is

1-(1-)2=1-(1- /k)2

Total Blocking probability is

PB=[1-(1-)2]s=[1-(1- /k)2]s

the
meaning
of k?

Probability graph

Remarks on PB of three-stage ~
The notation k=s/p represents either
space expansion (k>1) or
concentration (k<1).
Low PB requests small / k

In case of small , e.g. end offices and

PBX, k can be small. It means
concentration can be used in first stage.
In case of large , e.g. transit exchanges,
k should be large. It implies that
expansion should be used in first stage.

Nonblocking three-stage networks

C. Clos: pioneer researcher of multistage
nonblocking networks.
Clos networks
Multistage nonblocking and fully availabile
networks.
Much less switching elements are used than that
in single stage networks.

Design strategy
Providing adequate number of blocks in middle
stages. For three-stage networks, the value of s
should be large enough.

Three-stage nonblocking configuration

Worst situation for blocking
p-1 inlets in a block I in the first stage are busy;
p-1 outlets in a block O in the third stage are
busy;
The p-1 second-stage blocks, on which the p-1
outlets from block I are terminated on, are
different from the p-1 second-stage blocks from
which the links are established to the block O.
The free inlet of block I needs to be terminated
on the free outlet of block O.

How big should s be to avoid occurrence of

blocking?

Three-stage nonblocking configuration

Three-stage nonblocking configuration

The number of blocks required in the

second stage for nonblocking
operation is s=2(p-1)+1=2p-1.
The number of switching elements in
the nonblocking configuration is given
by
S=p(2p-1)r+(2p-1)r2+p(2p-1)r
=2N(2N/r-1)+r2(2N/r-1)
=(4N2-2Nr)/r+2Nr-r2

Three-stage nonblocking configuration

There exists an optimal value of r for
minimizing the value of S.

Let dS/dr=0, we have r2(N-r)=2N2.

For large values of N, we have N-rN.
Hence, r=(2N)1/2, p=N/r=(N/2)1/2
The minimum S is Smin=4N(2N)1/2

Switching elements advantage ratio

S in nonblocking single-stage network, N2
= S in nonblocking three-stage network, 4N(2N)1/2
See table4.4 on page 95.

4.9 n-Stage networks

Further reduction in the number of
switching elements are possible by
using even higher number of stages
than three.
Construction of multi-stage networks
By replacing the middle blocks with
three-stage network blocks continually,
any number of stages can be obtained.

Construction of five-stage ~

Assignments
Ex.11
Ex.15