1. What are the main driving applications for traffic growth in today's ...

somberastonishingAI and Robotics

Nov 13, 2013 (3 years and 8 months ago)



What are the main driving applications for traffic growth in today’s

Supercomputer Interconnections


Interactive TV

Telemedicine Applications

GRID Computing

Storage Area Networks

Distance Learning

Video Conferencing

Audio and Video Broa
dcasting etc.


Which technologies and components can be used to implement high
network elements? Only itemize the main classes of electrical,
optical and optoelectronic components and the applications in which they can
be used.

Electric componen


Integrated circuits






Transistors (MESFET, HEMT, HBT)

Optical components


Optical Fibres


Optical Couplers


Optical Isolators


Optical Filters

Optoelectronic components


Optical Sources (LEDs, LDs)




Optical Amplifiers


al Switches


What technologies and semiconductor materials are used to manufacture
integrated circuits for communications applications?



Bipolar (analogue)







What are the main trends in semiconductor electronics and in p
articular in
integrated circuits?

decreasing feature size

increasing chip area

increasing clock frequency

efficient micro structures

better design strategies and tools


How can integrated circuits be classified according to their flexibility and

Standard Circuits, such as a standard processor, are the most flexible integrated circuits. They are
normally produced in high quantity and have a relatively low price. Full
custom and semi
circuits are usually used to implement user specific

functions or customized blocks. Therefore they are
produced in a specific quantity and are more expensive.


Describe methods that can be used to increase performance of a CPU?


Application specific ISA (Instruction Set Architecture) extensions

ltiple ALUs and control units


Very Long Instruction Word (VLIW)


Memory hierarchy design


What is a content addressable memory (CAM)? Describe at least three
applications of CAMs.

CAM is an extension of ordinary memory (e.g. SRAM).

Read and write operations are performed as
usual. Content itself or parts of it function as key to search data. CAMs compare in parallel input
search data to all contents of memory and return the address of the data.
Tasks that benefit from
CAMs are searc
, such as the following applications

Routing address translation

time compression and encryption

Cache memory in microprocessors

Mapping tables and translation buffers in microprocessors

Pattern recognition in artificial intelligence robot
ic systems


Draw a block diagram of a basic CAM structure.


What is a ternary CAM (TCAM)? Highlight differences between basic (binary)
CAMs and ternary CAMs. What is the main networking application of TCAMs?

While binary CAMs store only two states (0, 1) t
ernary CAMs store an additional state (don’t care
state, X). When a don’t care is stored in the cell, a match occurs for that bit regardless of the search
data. TCAM devices are favoured by most network component and equipment vendors due to the fast
and d
eterministic lookup performance afforded by their use of massive parallelism.


What types of fibres are used in optical communications? How are the
transmission windows of standard single
mode fibres defined? Are there any
new fibre types and what are the t
rends in optical communications?

An optical glass fibre consists of a fibre core made of highly pure glass with a high refractive index. A
middle layer (cladding) is made of glass with a lower refractive index. An outer polymer jacket is added
to protect t
he fibre from damage.

There are three different kinds of optical fibre:

Multimode step

Multimode graded

mode (step

The attenuation characteristic offers three “windows”

or bands for transmission.

The first window (800
900nm) has

indeed a relatively high attenuation but allows

low cost optical
sources and detectors.

The second window (around 1310nm) offers zero dispersion on single
mode fibres. The attenuation is
lower than in the first window, but sources and detectors are
more costly.

The third window (1500
1600nm) has the lowest attenuation but sources and detectors are more
costly. Additionally the fibre disperses the signal in this band.

New types of fibres are AllWave (no attenuation peak around 1400nm) and Photonic Cry
stal Fibres
(PCF) that offer unique properties impossible to achieve in classical fibres.


How can chromatic dispersion in single
mode fibres be effectively
compensated? Describe briefly the nonlinear effects that occur in single

Chromatic disp
ersion describes the tendency for different wavelengths to travel at different speeds in a
fibre. In general, it consists of the material and waveguide dispersion.
The waveguide dispersion
components refer to the tendency that the effective refraction inde
x of a mode changes with
frequency. If the area occupied by the mode in relation to the fibre’s refractive index profile is large, the
change of the refractive index with frequency is also large. As the core diameters in single
mode fibres
are small, the w
aveguide contribution to the chromatic dispersion becomes comparable to the material
contribution in such fibres. The waveguide contribution shifts the zero
dispersion wavelength slightly
towards longer wavelengths. This is used to adjust the zero
on wavelength of the fibre by
adjusting fibre
design parameters such as core diameters and core
cladding index difference. By
using multiple cladding layers, so called dispersion
flattened fibres (DFF) can be designed.


Describe the main parameters and app
lications of optical filters. What
types of optical filters are used in c
ommunication networks?

Main filter parameters are:

Filter bandwidth

Centre wavelength

Insertion loss

Spectral shape

Contrast (ration between maximum values of transmission and reflect

Filters used in communication networks:

Perot filter (fixed and tuneable when mounted on piezoelectric crystals)

Diffraction Gratings

Fibre Bragg Gratings

Arrayed Waveguide Grating

optic tunable Filters

Dielectric Thin
Film Filters

ender Interferometer


Describe the differences between the gain
guided and the index
laser diode structure. What are the main differences between a Fabry
(FP) laser and a distributed feedback (DFB) laser? What structures can be used
to produce
short optical pulses?

The Fabry
Perot Laser is built up like a LED, but with a pair of mirrors at the ends. The mirrors are
necessary to create the right condition for lasing to occur. The active layer is very thin and the
refractive index difference betwe
en the material of the active layer and the surrounding material is not
great. Thus you don’t get lasing in the vertical (transverse) mode. Lasing in the lateral mode is
minimized by an anti
reflection material or just rough sides.
Lasing in the longitudin
al mode across the
full width of the device is a problem as the device will tend to produce many different localised areas of
lasing at different wavelengths. It is additionally a significant problem to guide the light into a fibre. To
get rid of this prob
lem gain
guidance and index
guidance have been developed.

guidance is achieved by limiting the area of the electrical contact on the surface of the device.
Power is delivered into the active layer in a long stripe. There will be
sufficient gain alon
g this path for
lasing to occur but outside the region there will not be sufficient gain to sustain lasing. Thus one gets
narrow beam of light issuing from the centre of the active region.

guidance is achieved by using a lower refractive index in the

active region than the material
surrounding it. In this situation the light is guided out of the cavity.


Draw the schematic of an erbium
doped fibre amplifier (EDFA). Are there
any other optical amplifiers that can be used in fibre communications?
ibe the operational principle of these optical amplifiers and in which
applications they can be used.

Doped Fibre Amplifier

An Erbium
doped fibre amplifier consists of a short section of fibre which has a small controlled
amount of Erbium added to

the glass. The Erbium ions are able to exist in several energy states. The
erbium ions are excited into a high
energy state by a pump
laser at an appropriate wavelength. These
ions in the excited state are then available for stimulated emission and amplif
y the signal.

Fibre Raman Amplifier

The relatively large bandwidth makes them attractive for fibre
optical communication applications.
Moreover due to the distributed amplification the signal to be amplified remains in the linear
amplification region, s
o that the impact of nonlinear effects and the accumulation of noise are reduced.

Semiconductor Optical Amplifier

SOAs are basically semiconductor laser diod
s with antireflection coated sides that prevent
oscillations in the cavity.

In communication sy
stems, SOAs are used for a wide spectrum of applications such as high
optical switching, wavelength conversion, in
line amplification for metro networks, per
amplification in
optical receivers, and optical amplification and reshaping (2R), & retiming

(3R) regenerators for long
haul transmission systems.


How does a thermo
optic switch work? What are micro
mechanical system (MEMS) switches? Why are interferometric structures used
for optical switches?

optic switches are interferometric s
witches (usually Mach
Zender interferometers) where the
phase in changed in the arms by changing the refractive index of the waveguide material by changing
the temperature. This is done by varying the voltage applied to thin metal heating electrodes.

hing speed is low at about several microseconds.

mechanical system (MEMS) switches consist of small mirrors that are moved as desired
for switching. They have low insertion loss, low crosstalk, are optically transparent, and polarization
ensity, and can scale up to high port counts. But they are often bulky and require a complex optical
alignment and assembly.


Describe evolution of interconnection technologies during the last


What types of interconnects do you know? How can inter
connects be
classified? Describe benefits and drawbacks of three typical interconnection

In general, interconnects can be divided into tree main categories:

module (backplane) interconnects

chip interconnects

path (in
chip) in

Typical interconnection solutions are:

Peripheral Component Interconnect (PCI, PCI
X, PCI Express)

Small Computer System Interface (SCSI, SAS, iSCSI)

HyperTransport (HT)

Rapid I/O


Common Switch Interface (CSIX)




Fibre Channel


Describe PCI Express layered architecture. What are the main features of
the three lower layers?

The PCI Express architecture is specified in layers.

The software layers will generate read and write requests that are transported by the tra
nsaction layer
to the I/O devices using a packet
based split
transaction protocol. The link layer adds sequence
numbers and CRC to these packets to create a highly reliable data transfer mechanism. The basic
layer consist of a dual
simplex channel that is
implemented as a transmit pair and a receive pair.


Describe the main features of Serial Attached SCSI (SAS). Draw a typical
configuration of a SAS system.

Serial Attached SCSI (SAS) is a point
point architecture in which all storage devices connect
ectly to an SAS port rather than sharing a common bus as traditional SCSI devices do. Point
point links increase data throughput and improve the ability to locate and fix disk failures. SAS inherits
its command set from parallel SCSI, frame formats from

Fibre Channel, and physical characteristics
from Serial ATA. SAS is defined for full
duplex operations. Point
point configurations provide for
high bandwidth, but require intermediary devices between initiator devices (or hosts) and target
devices (or
peripheral devices) to provide a topology where there may be more than two devices in a
system. Inexpensive expanders are the intermediary devices defined for SAS
. Using SCSI protocol
and architecture, it is possible to bridge from a SYS system to other sy
stems using Infiniband, iSCSI or
Fibre Channel, which also use the same SCSI objects.


What is Internet SCSI (iSCSI)? Describe encapsulation of storage data
into TCP/TP packets by using the iSCSI protocol. There are three options for
implementing iSCSI ho
st bus adapters (HBA). Describe these implementation
options. What option can provide the highest performance?

Internect SCSI (iSCSI) is an standard that defines the encapsulation of SCSI packets in TCP which
allows them to be routed using IP. The storage
data is encaplulated into TCP/IP packets as shown in
the picture.

There are three classes of iSCSI host adapters

iSCSI in Software: is a very economical, low
performance block I/O solution, iSCSI drivers can
be run on an off
shelf Ethernet card or in
terface. Software implementation of iSCSI offers
advantages in terms of cost
s. H
owever it must be balanced with the penalty it imposes in
performance and CPU overhead.

iSCSI in Software with TCP off
load: TC

incurs significant processing overhead on the h
CPU. Off
loading this

processing to a host network interface card frees host CPU cycles and
enables much higher performance solutions.

iSCSI in silicon with TCP off
load: High performance iSCSI adapters off
load both TCP and
iSCSI processing to the int
erface card. While this adds costs to the adapter, it provides both
speed iSCSI transport and minimal CPU overhead.


What is HyperTransport (HT)? Describe HyperTransport connections and
device configurations. Can large data packets be transmitted direc
tly over a
HyperTransport channel (without need for segmenting the packet)? What is the
difference between host
reflected routing and device
device routing in

HyperTransport defines direct connection to the CPU or several CPUs via high b
andwidth, very low
latency links. The HyperTransport daisy
chain topology includes a required “host” device, at least one
point or “cave”, optional “tunnel” devices that connect the link to other HyperTransport devices
and optional “bridge” devices tha
t interface with non
HyperTransport interconnect technologies.

allows transmitting packets of the size of 4
64 bytes directly in a single
HyperTransport data packet. If it is longer than 64 bytes it can be placed in a sequence of data
s by breaking it up into segments of 64 bytes.

In HyperTransport there are two different routing mechanisms specified:

Host reflected routing: requires that all traffic passes through the host in order to maintain PCI
compatibility. All traffic passes thro
ugh the host device of the daisy chain. Thus, in order to
replicate the ordering properties of PCI, all base HyperTransport transactions are reflected
through the host.

device routing: allows two devices to communicate directly, greatly reducing
traffic and off
loading the host from reflecting the traffic.


What are the main characteristics of RapidIO?

RapidIO architecture is an electronic data communication standard for interconnection chips on a
circuit board and circuit boards using a backp
lane. It is a packet
switched interconnect architecture
conceptually similar to IP. However it is designed to be used for the processor and peripheral interface
where high bandwidth and low latency are necessary. RapidIO is specified in a
layer architect

Logical layer: defines the overall protocol and packet formats. This is the information
necessary for end points to initiate and complete transactions

Transport layer: provides the necessary route information for a packet to move from end
to end point.

Physical layer: describes the device level interface specifics such as packet transport
mechanisms, flow control, electrical characteristics, and low
level error management.

This partitioning provides flexibility to add new transaction t
ypes to the logical layer without requiring
modifying the other layers.

The main characteristics of RapidIO are:

Packet switching: point
point interconnect to connect processors, co
processors, memory
and memory mapped I/O

Low overhead and low latency

mall silicon footprint: can be implemented in ASICs and even FPGAs

Software transparent: an extension of microprocessor bus that allows direct, physical memory
mapping of the entire machine

Reliable delivery of packets: error detection and recovery in hard

Layered architecture

Standard I/O technology


What are the main components of an InfiniBand subnet? Describe the
functions of a local (LID) and a global (GID) identifiers.

The InfiniBand architecture (IBA) is an industry
standard architecture for serve
r I/O and inter
communications. The elements of a subnet are endnodes, switches, links and a
Endnodes, such as hosts and devices, send messages over links to other endnodes. The messages
are routed by switches. Routing is defined, an
subnet discovery performed, by the Subnet Manager.
Channel Adapters (CAs) connect endnodes to links.

Every IBA subnet must contain a single master Subnet Manger, residing on an endnode o
a switch. It
discovers and initializes the network, assigning Loca
l IDs (LIDs) to all elements, determining path
maximum transfer units (MTUs), and loading the switch routing tables that determine the paths from
endnode to endnode. Each endnode has one or more CAs and each CA has one or more ports. Each
port has a unique

bit LID and at least one IPv6 address (Global ID, GID). Three types of quantities
are important to IBA addressing: LIDs, GUIDs and GIDs.

LIDs are subnet unique 16
bit identifiers used within a network by switches for routing.

GUIDs are Global unique ID
s are identifiers for elements in a subnet.

GIDs are 128
bit global identifiers used for routing across subnets.


What are the main applications for Common Switch Interfac
s (CSIX)?
Describe the main functions of the physical, interconnection, and logical
message CSIX levels. What is a CFrame?

CSIX is a detailed interface specification between port/packets processor logic and interconnect fabric
logic. It is a scalable parallel interface with separate data and control paths.
It is e generic specification

promote the deployment and development of highly scalable network switches.
There are three
levels in CSIX:

Logical or Message Level: ensures that data or control message protocol exchanged over the
interface are properly understood by each end and proper
ly processed by the appropriate

Interconnection Level: defines all the signals with specific functions, meanings, and bit widths,
input or output, signal handshake protocols, etc.

Physical Level: specifies the electrical characteristics such as vo
ltage levels, capacitance,
drive strengths, timings, etc.

A CFrame is the base information unit transferred between traffic managers and a CSIX fabric. It
consists of a base header, an optional extension header, and optional payload, optional padding bits
and a 16
bit vertical parity field. The payload is variable in length and is passed by CSIX fabric from
incoming traffic manager to the outgoing traffic manager.


What technologies can be used to implement high
capacity backplanes?

There are two types of ba
ckplanes: active and passive. Active backplanes contain bus control bridges;
however they do not contain processor complex components such as the CPU, chipset or cache.
Passive backplanes contain circuitry for bus connectors and, in some cases, buses and d

To avoid single point failure, switching function can be distributed over the line cards. Thus, in case
each card implements a switching element, and therefore, no active backplane is needed. A drawback
of this solution is that the line cards becom
e more complex and more costly. Advanced switching
should provide a standard backplane for interface for line cards, switch cards and control cards used
in communication system applications.


Describe the SONET/SDH interfaces developed by the Optical
etworking Forum (IOF). What are the main differences between System
Packet Interface (SPI) and DERDES Frame Interface (SFI)?

The OIF developed the System Packet Interface (SPI), the SERDES Framer Interfaces (SFI) and the
TDM Framer Interface (TFI).


is between the Physical Layer device(s) and the rest of the SONET/SDH System. This
interface separates the synchronous PHY layer from the asynchronous packet
based processing
performed by the higher layers. As such, the SPI supports transmit and receive d
ata transfers at clock
rates independent of the actual line bit rate. It is designed for the efficient transfer of both variable
sized packet and fixed
sized cell data.

The SFI defines an electrical interface between a SONET/SDH Framer and the high speed P
Parallel (SERDES) logic. This permits the SERDES and framer to be implemented in
different speed technologies, allowing a cost
effective multiple chip solution for the SONET/SDH
physical layer.

The TFI defines the backplane inte
rface (either electrical or optical) between a SONET based TDM
framer and the switch fabric. Traffic between the framer and the fabric is modelled after a
SONET/SDH frame, and operates at the STS
16 equivalent bit rate.


Describe functional layering
of a typical SONET/SDH framer device. What
kinds of interfaces are usually used as Layer 2 (L2) interface in SONET/SDH
framer devices?


Describe the media options for Ethernet systems. Which cable types and
coding methods are used in Ethernet? Describe the

coding methods used in
Gigabit Ethernet over twisted pair cables and optical fibres.

Gigabit Ethernet over twisted pair uses 4D PAM
5 Code.


What are the main functions of the Gigabit Ethernet physical layer?

The Physical Layer is subdivided into three s

Physical Coding Sublayer: auto
negotiation, synchronisation at byte level, carrier sense as
well as transmit and receive functions (including coder and decoder blocks

Physical Medium Attachment

Physical Medium Dependent


Draw a block diagram of th
R (10 Gigabit Ethernet) physical
coding sublayer (PCS). Describe the main functions of the blocks. What coding
scheme is used in 10GBASE


Itemize at least five Ethernet interfaces. Which interface replaces the
Attachment Unit Interface (AUI)
in Fast Ethernet? How is the physical layer
connected to Media Access Control (MAC) layer in Gigabit Ethernet?

Ethernet Interfaces:

MAU (Medium Attachment Unit)

MDI (Medium Dependent Interface)

AUI (Attachment Unit Interface)

TBI (Ten
Bit Interface)

Reduced Ten
Bit Interface)

MII (Medium Independent Interface)

GMMI (Gigabit Medium Independent Interface)

RGMII (Reduced Gigabit Medium Independent Interface)

GBIC (Gigabit Interface Converter)

XAUI (10 Gigabit Attachment Unit Interface)

XSBI (10 Gigabit S
Bit Interface)

XGMII (10 Gigabit Medium Independent Interface)

The Medium Independent Interface (MII) replaces the Attachment Unit Interface (AUI) in Fast
Ethernet. The physical layer is connected to the Media Access Control via the Gigabit Medium
ndependent Interface (GMII) in Gigabit Ethernet and via the 10 Gigabit Medium Independent Interface
(XGMII) in 10 Gigabit Ethernet respectively.


What are the main features and applications of the 10 Gigabit Attachment
Unit Interface (XAUI)? What is the max
imum distance possible by using XAUI

XAUI is a full duplex interface that uses four self
clocked serial differential links in each direction to
achieve 10 Gbit/s data throughput.


Describe the main features of the Universal Test & Operations PHY
Interface for ATM (UTOPIA). What are the differences between four levels of
UTOPIA interface?

UTOPIA is defined by the ATM Forum to provide a standard interface between ATM devices and ATM
PHY or SAR (segmentation and Reassembly) devices.


Describe the tran
smission hierarchy and the concept of flow control
used in Fibre Channel (FC). How many classes of services are there in FC?
Which components can be used in FC systems?

Basically every byte of data that is to be transmitted is first converted into a 10

value called
Transmission Character. All information in Fibre Channel is transmitted in groups of four Transmission
Characters called Transmission Words. Together with 36 bytes of overhead and several full
Transmission Words are transmitted in a FC frame

(word in the analogy)
. A series of one or more
related frames is called a Fibre Channel Series

(sentence in the analogy)
. A Fibre Channel Exchange
(conversation in the analogy)
is a series of one or more non
concurrent sequences between two ports

A devic
e can transmit frames to another device only when the other device is ready to accept them.
Before the devices can send data, they must login to each other. Thereby credit is established which
refers to the number of frames a device can receive at a time.
After credit runs out, no more frames
can be transmitted until the destination device grants more credit. In FC two types of flow
control are
used: buffer
buffer and end

There are 6 numbers of classes (communication strategies) in FC, depending

on the type of data to be

Components that can be used in FC systems are:

FC Hubs

Switches & Directors

Host Bus Adapters (HBA)