OPTICAL BURST SWITCHING

OPTICAL BURST SWITCHING

ABSTRACT:

Optical burst switching is a promising solution for all optical WDM

networks It combines the benefits of optical packet switching and wavelength routing

while taking into account the limitations of current all optical technology In OBS the user

data is collected at the edge of the network, sorted based on destination address,and

grouped into variable sized bursts Prior to transmitting a burst, a control packet is created

and immediately send toward the destination in order to setup a buffer less optical path

for its corresponding burst After an offset delay time, the data burst itself is transmitted

without waiting for positive acknowledgement from the destination node the OBS

framework has been widely studied in the past few years because it achieves high traffic

throughput and high resource utilization .In this paper we present in a systematic way the

main objectives of OBS design parameters and the solutions that have been proposed

INTRODUCTION:

Optical communication has been used for a long time and it very

much popular with the invention of wavelength-division multiplexing(WDM) Current

WDM works over point-to-point links,where optical-to-electrical-to-optical(OEO)

conversion is required over each step The elimination of OEO conversion in all optical
networks(AON) allows for unprecedented transmission rates AON’s can further be

categorized as wavelength-routed networks(WRNs).,optical burst switched

networks(OBSNs),or optical packet switched networks(OPSNs).Now we discuss here

about optical burst switching(OBS)

In optical burst switching(OBS) data is transported in variable

sized units called bursts Due to the great variability in the duration of bursts the OBS

network can be viewed as lying between OPSNs and WRNS That is, when all burst

durations are very short,equal to the duration of an optical packet,OBSN can be seen as

resembling an OPSN On the other hand,when all the burst durations are extremely long

the OBSN may seem resembling a WRN In OBS there is strong separation between the

data and control planes,which allows for greater network manageability and flexibility In

addition its dynamic nature leads to high network adaptability and scalability,which

makes it quite suitable for transmission of bursty traffic

In general,the OBS network consists of interconnected core

nodes that transport data from various edge users The users consist of an electronic router

and an OBS interface, while the core OBS nodes require an optical switching matrix,a

switch control; unit and routing and signaling processors OBS has received considerable

attention in the past few years and various solutions have been proposed and analyzed in

an attempt to improve it’s performance Here we describe the various OBS architectures

by grouping the material logically per OBS design parameter

BURST AGGREGATOIN:
 OBS collects upper layer traffic and sort it based on destination

addresses and aggregate it into variable size bursts The exact algorithm for creating the

bursts can greatly impact the overall network operation because it allows the network

designers to control the burst characteristics and therefore shape the burst arrival traffic

The burst assembly algorithm has to consider a preset timer and maximum and minimum

burst lengths The burst aggregation algorithm may use bit-padding ,the differentiation of

class traffic , create classes of service by varying the preset timers and

maximum/minimum burst sizes

One of the most interesting benefit of burst aggregation is it shapes the

traffic by reducing the degree of self-similarity,making it less bursty in comparison to the

flow of the original higher-layer packets Traffic is considered bursty if busy periods with

a large of arrivals are followed by long idle periods The term self-similar traffic refers to

an arrival process that exhibits burstiness when viewed at varying time

scales:milliseconds,seconds,minutes,hours even days and weeks Self-similar traffic is

characterized by longer queuing delays therefore degrades network performance

Therefore reducing self-similarity is a desirable feature of the burst assembly process and

concluded that traffic is less self-similar after the assembly.

CONNECTION SETUP MECHANISMS:

signaling, routing, and wavelength allocation Signaling is used to setup and tear down the

connections for the bursts Routing is used to decide the path of a burst through the OBS

network Wavelength allocation is used to determine on which wavelength to transmit the

burst

SIGNALING FOR OBS:

Signaling specifies the protocol by which the OBS nodes

communicate connection requests to the network and its operation determines whether or

not the resources are utilized efficiently

Distributed Signaling With One-Way Reservation:

In this prior to transmitting a burst a user transmits a control

packet to its ingress OBS node This control packet information about the corresponding

burst and is electronically processed by the ingress OBS node and all the subsequent

nodes along the path of the destination user The control packet is transmitted in an out of

band control channel which may be a wavelength dedicated to signaling or a separate

electronic control network such as an IP or asynchronous transfer mode(ATM) network

In either case the separation of control and data in both time and physical space is one of

the main advantages of OBS It facilities flexibility in the user data format and rate

because the bursts are transmitted entirely over an optical signal and remain transparent
throughout the OBS network The burst itself is transmitted after a delay known as the

offset without waiting for a positive acknowledgement that the entire path has been

successfully established

Due to the one-way reservation scheme burst loss may occur

in an OBSN because the control packets may not succeed in reserving resources at some

of the intermediate OBS core nodes In addition burst loss is possible if the control

channel itself suffers from congestion or other failure Because of these reasons the burst

loss probability is an important performance measure of an OBS architecture Despite the

fact that burst loss is possible in OBS the proposed architectures do not implement

retransmission of lost bursts One reason is the high data rate which makes it

unmanageable to keep copies of all previously transmitted at the OBS edge nodes

Therefore retransmission of lost bursts in an OBS network is left as a responsibility of the

higher layer protocols It is also possible that an application may tolerate burst loss in

which case there is no need of retransmission

Centralized Signaling With End-to-End Reservation:

Contrary to the more common one way signaling this centralized

signaling method termed wavelength routed optical burst switching (WR-OBS) which

utilizes an end to end resource reservation procedure In this design there is a centralized

request server responsible for resource scheduling of the entire OBS network When an

OBS ingress node receives a setup request from a user it sends a control a packet to the
centralized scheduler where it is queued up based on the destination address This

centralized server has global knowledge of the state of the OBS switches and wavelength

availability along all the fiber links The responsibility of this central server include

processing incoming control packets, determination of routes of the required destinations

and assignment of available wavelength along each link The central server processes the

control packet and sends a positive ACK to the OBS user upon receipt of which the node

transmits the burst

ROUTING:

The routing of a burst through an OBS network can be done on a hop by

hop basis as in an IP network using a fast table lookup algorithm to determine the next

hop Another approach is to use the multi protocol label switching(MLPS) The MLPS

idea is to assign control packets to forward equivalent classes at the OBS user’s in order

to reduce the intermediate routing time to the time it takes to swap the labels A third

approach is to use explicitly precalculated setup connections which can be established via

Constraint Based Route Label Distribution Protocol(CR-LDP) or Resource Reservation

Protocol with Traffic Engineering(RSVP-TE) Explicit routing is very useful in a

constraint based routed OBS network where the traffic routes have to meet certain Qos

metrics such as delay, hop count, bit error rate or bandwidth In addition in order to deal

with node or link failures OBS routing should also be augmented with fast protection and

restoration schemes unfortunately this is a weak point for explicit routing schemes
because sometimes the routing tables can become outdated due to the long propagation

time until a failure message reaches all of the OBS nodes

Wavelength Allocation: With or Without Conversion

In an OBS network with no wavelength converters the entire path from source to

destination is constrained to use a single wavelength. The other possibility is an OBS

network with a wavelength converter capability at each OBS node. In this case if two

bursts contend for the same wavelength on the same output port the OBS node may

optically convert one of the signals from an incoming wavelength to a different outgoing

wavelength In addition the conversion capability at an OBS node can be classified further

as full or sparse In the former case there one converter per each wavelength whereas in

the latter case the number of converters is less than the total number of wavelengths

Wavelength conversion is a desirable characteristic in an OBS network as it reduces the

burst loss probability However it may not necessarily be a practical assumption since all

optical converters are still an expensive technology Another important thing about OBS

wavelength allocation scheme is the fairness achieved between the successful

transmission of bursts over long and short paths The fairness issue is inherent to all

optical networks not just OBS networks and it is due to the fact that it is easier to find

free wavelengths along all of the links of a short path than it is for a longer one Therefore

the proposed all optical architectures should consider heuristics that try to improve the

fairness among the connections with different number of hops

An OBS user first transmits a control packet and

after an offset time it transmits the burst This offset allows the control packet to reserve

the needed resources along the transmission path before the burst arrives Furthermore the

OBS nodes need this offset time to setup their switching 0fabrics so that the data burst

can “cut through” without the need for any buffers Ideally the offset estimation should be

based on the number of hops between source and destination and the current level of

congestion in the network Obviously an incorrect offset estimation would result in data

loss because the burst may arrive at an OBS node before the optical cross connect has

been completely set up Therefore determining this offset is a key design feature of all

OBS networks and it’s effectiveness is measured in terms of burst loss probability There

are variations in the OBS literature on how exactly to determine the pretransmission

offset time and how to reserve the needed resources at he core OBS nodes Despite their

differences however all of the proposed OBS architectures have dynamic operation which

result in high resource utilization and adaptability

!.Fixed Offsets:

2.Statistical Offsets

3.WR-OBS Offsets:

SCHEDULING OF RESOURCES :(RESERVATION AND RELEASE):

nodes schedule their resources based on the included information The proposed OBS

architectures differ in their resource(wavelength) reservation and release schemes These

schemes are classified based on the amount of time a burst occupies a path inside the

switching fabric of an OBS node In explicit setup a wavelength is reserved and the

optical cross connect is configured immediately upon processing of the control packet In

estimated setup the OBS node delays reservation and configuration until the actual burst

arrives The allocated resources can be released after the burst has come through using

either explicit release or estimated release In explicit release the source sends an explicit

trailing control packet to signify the end of a burst transmission In estimated release an

OBS node knows exactly the end of the burst transmission from the burst length and

therefore can calculate when to release the occupied resources based on this classification

the following four possibilities exist they are explicit setup/explicit release explicit

setup/estimated release, estimated setup/explicit release and estimated setup/estimated

release

The burst assembly strategy implemented at the OBS users also dictates

how resources are reserved and released in the OBS network The two architectures

explicit setup/explicit release and explicit setup/estimated released schemes were used by

the Jumpstart project. The other two schemes were disregarded because of their necessity

for a scheduler at each node The Jumpstart signaling protocol however is designed to be

implemented mostly in hardware and does not use a scheduler One more OBS resource

scheduling scheme is Horizon This scheme can be classified as explicit setup/estimated

therefore the scheduler can maintain a deadline(horizon) when each resource will be

freed and available for future scheduling This scheme is categorized as explicit setup

because as soon as the control packet arrives at an OBS core node a wavelength is

immediately scheduled for the future burst arrival The horizon scheme is practical and

simple and its look ahead resource management minimizes the wasteful gap between

reservation time and the actual burst arrival

Variations On Burst Dropping:

Most of the OBS literature specifies that if all the

resources are occupied at the moment of a burst arrival the entire data burst is lost An

interesting OBS variation designed to reduce the probability of burst losses is presented

here It is built on the JET architecture and combines burst segmentation with deflection

routing Specifically in this OBS model each burst is divided into multiple segments in the

case of resource contention instead of dropping the entire burst either the head or tail

segment is deflected to an alternate route to the destination There were two ways to

implement this which are segment first or deflect first In the former the lengths of the

currently scheduled burst and the new contending burst are compared the shorter one is

segmented and its tail deflected In the deflect first policy the contending burst is deflected

if the alternate port is free However if the port is busy similar to the segment first policy
the lengths of the currently scheduled and contending bursts are compared and the tail of

the shorter one is dropped

Classes Of Traffic:

It is desirable for an OBS architecture to support different classes of

traffic in the user plane One reason is that applications such as video cannot tolerate long

queuing delay and therefore may need to be given higher priority than regular data traffic

In addition in order to ensure proper operation OBS protection and restoration traffic

must also be given priority over regular user data Specifically in an OBSN filtering of

upper layer data and assignment of priorities to bursts will occur at the edge of the

network during the burst assembly process Therefore in order to minimize the end to end

delay of high priority traffic the burst assembly algorithm can vary parameters such as

preset timers or maximum/minimum burst sizes However selecting the values for these

parameters is a difficult task because of the throughput interdependence some of the

proposed solutions

1.Classes Based On Extended Offsets:

2.Classes Based On Priority Queues:

3.Classes Based On the Optical Signal Properties and Pre-emption:

OBS Ring Networks:

review of OBS we now discuss the operation of ring networks. OBS ring architecture

consists of ‘n’ OBS nodes connected by optical fibers which support ‘n+1’ wave lengths.

Each of the OBS nodes has a fixed transmitter set to one of the ‘n’ wave lengths

and a tunable receiver. So that it can receive bursts along the transmission wave lengths

of the other nodes. In addition each of the OBS nodes is having a secondary pair of fixed

transmitter and a receiver set to the separate control wave length in order to communicate

control information along the ring. In this architecture it is possible for two OBS nodes to

send bursts over lapping in time towards the same destination node consequently these

bursts will contend for the tunable receiver of the destination node and one of them will

be dropped.

Conclusion:

In this we survey the OBS technology, a rapidly growing solution for all optical

WDM networks. We have described the various OBS design characteristics such as

connection establishment mechanism, offset time, scheduling of resources, aggregation

and loss of bursts, implementation of classes of traffic and additional multicast capability.

We present the ideas of deflection routing partial bursts dropping and fiber delay lines

because of their potential to lower the bursts loss probability in an OBS network. With

respect to the current state of technology OBS combines the best features of both circuit

switching and packet switching.

Source Node 1 Node 2 Destination

Off set

Processing
time
Switch
setup time

users
users
users

OBS node

WDM fiber line

users users
Users

The OBS network architecture