ECE 542

Telecommunications Engineering
Lecture 3
Switching and switches: Space and Time.

What you should know:

• Types of Switches
• Operation of the different switches
• Limitations of the switches
• Exchange stations

Switches
The primary function of a network is the interconnection of devices and the
sharing of resources (data) between these interconnected devices. When only
one or two devices are connected, this process is very straight forward and there
can simply be a direct single link (connection) between them. However, this is
not usually the case as a typical network consists of several devices. The problem
of how these devices can be connected together and have them sharing
information now arises. This can be solved by using switches. A switch is a
device capable of creating temporary connections between two or more
devices linked to it. Several switches are used in a switched network where they
can be referred to as nodes. In a switched network, some switches are
connected to printers, computers, telephones and or networking devices (all
known as end devices) while others are used for routing. In the figure below, A,
B, C, D… are the end devices while I, II, III… are the switching devices

1|Page
Structure of a Switch
The switches used for circuit switching and packet switching are different. This
section will look at the features of these respective switches.
Structure of Circuit Switches
Two categories of switches are used in circuit-switched networks:
• Space-division switch
• Time-division switch
Space-division Switch:
In these kinds of switch, there is a spatial separation of the paths in the circuit.
This implies that there is a physical separation of the signal paths from each
other. Each connection requires the establishment of a physical path through
the switch dedicated solely to the transfer of signals between the two
endpoints. The original intention of this switch was for use in analogue networks
but it is currently also being used for digital networks. It has evolved over the
years and this evolution can be seen in the switches below.

Crossbar Switch:
A crossbar switch connects n inputs to m outputs in a grid configuration. It
employs electronic micro-switches (transistors) at each crosspoint to achieve
this.

2|Page
To connect n inputs to m outputs using a crossbar switch needs n x m crosspoints
ie, if you’re trying to connect 2000 inputs to 2000 outputs, the implication is that
you would require a switch with 4,000,000 crosspoints, this is obviously going to
be a physically large switch hence impractical.
There are a number of limitations to the crossbar switch:
• Cost. Because the number of crosspoints increases with the square
of the number of attached stations. This is noticeable for large
switches.
• Multiple points of failure. The loss of a crosspoint prevents
connection between the two devices whose lines intersect at that
crosspoint.
• Inefficient utilization of the crosspoints. Even when all of the
attached devices are active, only a small fraction of the crosspoints
are engaged. Statistics show that, in practice, fewer than 25
percent of the crosspoints are in use at any given time. The rest
are idle.

Multistage Switch:
In order to overcome the limitations of the crossbar switch, the multistage
switch was developed. What the multistage switch does is to combine crossbar
switches in several stages (usually 3).

3|Page
In a single crossbar switch, only one row or column (one path) is active for any
connection. So we need N × N crosspoints. If we can allow multiple paths inside
the switch, we can decrease the number of crosspoints. Each crosspoint in the
middle stage can be accessed by multiple crosspoints in the first or third stage.
The advantages of this multistage switch are:
• The number of crosspoints is reduced, increasing crossbar
utilization.
• There is more than one path through the network to connect two
endpoints, increasing reliability.
Multistage switches are designed using the following steps:
• We divide the N input lines into groups, each of n lines. For each
group, we use one crossbar of size n × k, where k is the number of
crossbars in the middle stage. In other words, the first stage has
N/n crossbars of n × k crosspoints.
• We use k crossbars, each of size (N/n) × (N/n) in the middle stage.
• We use N/n crossbars, each of size k × n at the third stage.
We can calculate the total number of crosspoints as follows:

4|Page
 𝑁 𝑁 𝑁 𝑁 𝑁 2
(𝑛 × 𝑘) + 𝑘 ( × ) + (𝑘 × 𝑛) = 2𝑘𝑁 + 𝑘 ( )
𝑛 𝑛 𝑛 𝑛 𝑛

This is smaller than the number of crosspoints in a single-stage switch which is

N2.
Example: Design a three-stage, 200 × 200 switch (N = 200) with k = 4 and n = 20.
In the first stage we have N/n or 10 crossbars, each of size 20 × 4.
In the second stage, we have 4 crossbars, each of size 10 × 10.
In the third stage, we have 10 crossbars, each of size 4 × 20.
The total number of crosspoints is 2kN + k(N/n)2, or 2000 crosspoints. This is 5
percent of the number of crosspoints in a single-stage switch (200 × 200 =
40,000).

The multistage switch in the example above has one drawback—blocking during
periods of heavy traffic. The whole idea of multistage switching is to share the
crosspoints in the middle-stage crossbars. Sharing can cause a lack of
availability if the resources are limited and all users want a connection at the
same time. Blocking refers to times when one input cannot be connected to an
output because there is no path available between them—all the possible
intermediate switches are occupied. For instance in the multistage switch
shown above, the heavier lines indicate the lines that are already in use. In this
state, input line 10, for example, cannot be connected to output line 3, 4, or 5,
5|Page
even though all of these output lines are available. This can be solved (made
non-blocking), albeit with increasing cost, by increasing the number or size of
the intermediate switches. Blocking is not an issue in single-stage switching
where there is always an available path to connect one input to an output.

Clos Criterion
Even though multi stage switches solved the problems of underutilisation of the
crossbars, there was still the issue of blocking. Because of this, French researcher
at Bells laboratory, Charles Clos, came up with the formulas below which when
adhered to sought ensure that there are enough paths to access all intended
devices. He was able to determine the number of switches needed at the
intermediate stage of a multi stage switch in order to eliminate the problem of
blocking.
n = (N/2)^1/2
k ≥ 2n – 1
Total number of crosspoints ≥ 4N[(2N)^1/2 – 1]
If the switch in the previous example (200 X 200 switch) is redesigned using Clos
criteria we would have:
n = (200/2)^1/2 = 10
k = 2n – 1 = 19
In the first stage we have 200/10 or 20 crossbars each with 10 x 19 crossbars (n
x k)
The second stage has k = 19 cross bars each with 20 x 20 crosspoints (N/n x N/n)
3rd stage has 200/10 or 20 crossbars each with 19 x 20 crossbars (k x n)
The total number of crosspoints is therefore: 20(10 x 19) + 19(20 x 20) + 20(19 x
20) = 15,200
It is clear that for a nonblocking situation, there has to be 15,200 cross points
instead of the 2000 earlier calculated when Clos criterion was not considered.
This can easily be verified using the formula: 4N[(2N)^1/2 – 1]

6|Page
Examples of Crossbar Switch
Crossbar switches have been used telecommunication exchanges. Examples
include the now decommissioned One Electronic Switching Systems (1ESS) and
its upgrade the 1AESS which was removed from service around June 2017.
They have also been used in data centre networking in applications such as their
use in the IBM System x3530 M4 server where they provide high-speed data
connections between processors, memory, and I/O subsystems.
In high performance computing, crossbar switches have been used as integral
part of high-performance computing (HPC) systems. Crossbar switching
technology can provide the low-latency, scalable interconnection required in
these systems. The Cray XT5 HPC system is one example that employed a
crossbar switch to interconnect its processors and memory banks to ensure
optimal performance.
AMD also produces programmable FPGA based crossbar switches.
Advantages of using a crossbar switch?
Some advantages of using a crossbar switch include:
• High switching capacity
• Simultaneous connections without interference
• Minimal signal attenuation or loss
• Scalability
• Fast and efficient signal routing
• Improved reliability
Common applications of crossbar switches include:
• Telecommunication systems for routing telephone calls and data traffic
• Data centre networks for connecting servers and storage devices
• Integrated circuits and multiprocessor systems for interconnecting
various components
• Video and audio routing in broadcasting and production facilities

Time-Division Switch:
Time-division switching uses time-division multiplexing (TDM) inside a switch.
TDM had enabled an advancement in the inefficient switching method which is

7|Page
provided by Space-division switching. Most of the modern switches use time-
division techniques to establish and maintain “circuits”. This technique involves
the partitioning of a lower-speed bit stream into pieces that share a higher-
speed stream with other bit streams. The individual pieces, or slots are
manipulated by control logic to route data from input and output. The most
popular technology is called the time-slot interchange (TSI).

Time-Slot Interchange (TSI):

The figure below shows a system connecting 4 input lines to 4 output lines. In
the figure, each input line wants to send data to an output line according to the
following pattern: (1→3), (2→ 4), (3 →1), and (4 → 2).
The figure combines a TDM multiplexer, a TDM demultiplexer, and a TSI
consisting of random access memory (RAM) with several memory locations. The
size of each location is the same as the size of a single time slot. The number of
locations is equal to the number of inputs. The RAM fills up with incoming data
from time slots in the order received. Slots are then sent out in an order
decided by the control unit.
TSI is a simple, effective way to switch TDM data. However, the size of such a
switch, in terms of the number of connections, is limited by the amount of
latency that can be tolerated. The greater the number of channels, the greater
the average delay that each channel experiences.

Comparisons between space and time switching gives rise to some interesting
facts. While space division has an advantage of being instantaneous, when the
operation of TSI is considered, it has a disadvantage of having delays because of

8|Page
the processing of each connection which is involved. Each time slot must be
stored by the RAM, then retrieved and passed on.
Limitations of TSI
The memory of the switch has to be accessed 2 times during the operation of
the switch; first during the write operation and second during the read
operation. If the time taken to access the memory is denoted as T, then time
taken during the operation is 2T (for read and write operations). If there are n
connections in the switch, then the operating time, t, for the connections is
given as:
t = 2Tn
hence n =t/2T
t and n are usually derived from a higher level system design of the switching
fabric. The implication of this is that the technology which results in T
determines n for a given t. T also limits t for a given n. Because of these
limitations, real life switches cannot be of any random configuration (ie large n
or small t).
Higher speed switches can have T halved by utilizing a more expensive 2-port
RAM where read and write can be executed at the same time. The switch must
still determine which takes precedence when there is an attempt to have read
and write actions at the same time.

Time-Space-Time (TST)/Time-Multiplexed (TMS) Switches

These refer to a combination of space-division and time-division technologies in
order to take advantage of the best of both. Combining the two results in
switches that are optimized both physically (the number of crosspoints) and
temporally (the amount of delay). Multistage switches of this sort can be
designed as time-space-time (TST) switches.
The figure below shows a TST switch which has two time stages and a single
space stage. Instead of one-time division switch, the inputs are grouped into 3
groups of 4 inputs and are directed to three time-slot interchanges. The effect
is that the delay is one-third of what would result from using one time-slot
interchange to handle all 12 inputs.

9|Page
The last stage is a mirror image of the first stage. The middle (second) stage is a
space division switch (crossbar) that connects the TSI groups to allow
connectivity between all possible input and output pairs.
This type of switching can also be implemented by using digital selectors (see
figure below). The selector (SEL) device selects an input line based on a
channel assignment provided from a store controlled by a time-slot counter.
To reduce or eliminate blocking, multiple stage networks can be built by
concatenating TMS (S) and TSI (T) stages.

Structure of Packet Switches:

A switch used in a packet-switched network has four components:
• Input ports

10 | P a g e
• Output ports
• Routing processor and
• Switching fabric

Input ports:
This performs the physical and data-link functions of the packet switch.
The bits are constructed from the received signal. The packets are
removed from the frame. Error detection and correction is also done at
this point after which the packet is now ready to be routed by the network
layer. There is also some buffering done at the input port to hold the
packets before they are directed to the switching fabric.

Output ports:
This performs the same functions of the input ports but in the
reverse order.

11 | P a g e
Routing Processor:
This performs the network layer function in the switch. It uses the
destination address to find the address of the next hop and, at the same
time, the output port number from which the packet is sent out. This
process is often referred to as table lookup because the routing processor
searches the routing table. In order to increase speed, this process is being
moved to the input ports in newer packet switches.
Switching Fabric:
The most difficult task in a packet switch is to move the packet from the
input queue to the output queue. The speed with which this is done
affects the size of the input/output queue and the overall delay in packet
delivery. In the early days, packet switches were dedicated computers
which the input ports store the packets in its memory and the retrieval
was done by the output ports. This memory was the switching fabric.
Today, packet switches are specialized mechanisms that use a variety of
switching fabrics such as:
• The crossbar switch which was discussed earlier
• Bayan switch - A banyan switch is a multistage switch
with microswitches at each stage that route the
packets based on the output port represented as a
binary string. For n input and n outputs we have log2n
stages with n/2 microswitches at each stage. It has
three stages: Left bit, middle bit and right bit.

12 | P a g e
 • Batcher-Bayan switch – An improvement on the Bayan
switch. Due to the possibility of internal collision in
Bayan switches, the Batcher-Bayan switch aims to
solve this my first sorting arriving packets based on
destination port. It does this by using a ‘trap’ module
which prevents duplicate packets from passing to the
Bayan switch.

Exchange stations: Public and Private

Public exchange

13 | P a g e
An exchange station is a system of electronic components that enable the
establishment of telephone calls connection.
A public exchange is a telephone exchange that services the general public. It
enables the interconnection of telephone subscribers within a geographical
region. It also facilitates the interconnection of subscribers between regions.
The aggregation of worldwide public exchanges is known as the public switched
telephone network.
Private exchange
This is an internal exchange which serves a single organization. It is used for
communication between staff of the organisation. It has no connection with a
public exchange. Some types of Private exchange (PX) include the private
automatic branch exchange (PABX) which provides for the transmission of calls
internally and to and from the public telephone network, private telex exchange
(PATX) which provides telex service within and outside and organisation, private
manual branch exchange (PMBX) which is a manual private exchange and the
private digital exchange (PDX) which uses digital transmission techniques.

14 | P a g e