S The Burst Loss Ratio And Packet Loss Rate

Published Date: 02 Nov 2017

In this chapter we list the conclusions, recommendations and future scope of the research.

8.1 CONCLUSIONS

OBS is considered as an optical networking technique that allows dynamic sub-wavelength switching of data. The purpose of OBS is to dynamically provision sub-wavelength granularity by optimally combining electronics and optics. OBS considers sets of packets with similar properties called bursts. OBS granularity is finer than optical circuit switching. OBS provides more bandwidth flexibility than wavelength routing but requires faster switching and control technology. OBS can be used for realizing dynamic end-to-end all optical communications.

OBS is viewed as a compromise between the yet unfeasible full optical packet switching and the mostly static optical circuit switching. It differs from these paradigms because OBS control information is sent separately in a reserved optical channel and in advance of the data payload. These control signals can then be processed electronically to allow the timely setup of an optical light path to transport the soon-to-arrive payload.

In this thesis, we concentrated on different performance related issues arising in the deployment of OBS networks such as network scheduling problem, burst assembly scheme to reduce burstification delay, wavelength reservations techniques for efficient utilization of bandwidth as well as network resources and contention resolution schemes, in order to reduce congestion and burst loss at the higher layers to support various internet and multimedia applications.

The following are the conclusions based on the results achieved in the individual chapters of the thesis:

Considering the issues related to the performance of OBS networks by using various Data Burst Scheduling techniques, we presented an efficient scheduling technique called MLAUC-VF in OBS networks. The proposed burst scheduling technique efficiently utilizes the voids created by previously scheduled bursts, resulting in high bandwidth utilization. The performance of proposed algorithm has been compared with existing Horizon Scheduling algorithm. The proposed algorithm uses bandwidth more efficiently as numbers of void checks are enhanced and thus improving the Burst loss ratio and packet loss rate parameters as compared to existing state-of-the-art scheduling algorithms. The proposed scheduler is able to produce a viable burst scheduling while having comparable processing speed as the well-known Horizon scheduler. Without loss of generality, we evaluated the end-to-end performance of this algorithm by varying the offset between the burst header and data burst. The results obtained show that the MLAUC-VF algorithm is well suited to high-performance emerging optical networks.

We have also investigated and presented burst assembly scheme for efficient network utilization and bandwidth in OBS networks. A congestion based hybrid burst assembly technique (VT-BA) which varies itself according to the traffic conditions in order to improve the OBS network performance has been proposed. The proposed burst assembly technique varies the time within fixed size threshold limits of the virtual queues at optical nodes considering the link congestion levels. We use the byte loss rates on the outgoing links of the optical nodes to decide on the threshold values (LT and LS) of related virtual queues. These parameters are varied with the burst loss rates on the corresponding links. Bursts are forced to have one of the following burst time threshold values in order to make implementation practical: LTlong, LTmedium, LTshort. We evaluate the performance of the proposed algorithm in terms of byte loss rate, average end-to-end delay and average burstification delay under various traffic classes. The results obtained are compared with those of the existing hybrid burst assembly algorithm. It is observed that the proposed scheme outperforms the hybrid burst assembly with respect to byte loss rate and delays.

Further, in order to find the optimal combination of burst assembly and burst scheduling techniques for superior performance of the OBS networks, the performance of proposed Optical Burst Assembly Technique is experimentally investigated under self-similar traffic and traditional scheduling techniques. To maximize throughput, we considered four different burst scheduling approaches and evaluated their performance for their compatibility with newly introduced Burst Assembly technique VT-BA, to attain more efficient OBS network. Simulation results reveal that the proposed burst assembly algorithm performs well amid various scheduling techniques and gives better compatibility with VT-BA as compared to their other counterparts for most of the parameters under consideration. We observe that CTBR and Min-SV scheduling techniques performs substantially better in terms of Traffic sent, Burst loss rate and Packet loss rate. Although, in the OBS literature, burst loss rate is a commonly employed metric, it is observed that the burst loss rates obtained are different than the corresponding packet loss rate. Such a result is caused by variable burst sizes. Moreover, the algorithm is simple to implement making them viable alternative for the design and implementation of burst assembly units in next-generation transport networks.

We have also investigated the performance of OBS networks using different signaling protocols. We presented an efficient multiple time slots (MTS) signaling protocol, based on time division multiplexing, for optical burst switching network has been proposed. The burst assembly process is also combined with the two-way reservation protocol to reduce the burstification delay, packet loss, end-to-end delay and at the same time utilizes the bandwidth more efficiently. The MTS signaling framework has been investigated to support entire OBS network services with diverse bandwidth granularities. The proposed signaling protocol reduces the burst loss probability as compared to conventional wavelength reservation approach TAW. In addition to the burst loss probability, the end to end delay has been evaluated and it is seen that the end to end delay of proposed technique is better than two-way signaling technique TAW. Furthermore it is observed that the proposed MTS results in improving the bandwidth utilization and provide the most flexible solution for the next generation optical networks. Results show that the proposed MTS is an improved, capable and more efficient solution for OBS networks.

Further to improve the burst loss probability, a blended signaling system which is known as the centre-node-reservation protocol (CNRP) for optical burst-switch networks has been proposed. This proposed protocol can provide both the two way and one way based signaling techniques which reduces the burst loss probability. In the proposed protocol, reservation of channels is acknowledged at a centre node instead of destination node, before transmitting the burst into the network in a simultaneous forward direction. The performance of the proposed scheme is investigated for benefits and further compared against existing Just-enough-time (JET) scheme through simulations. The burst loss probability has been evaluated and it is seen that the burst loss probability of proposed technique is better than one-way signaling technique JET. Furthermore it is observed that the proposed CNRP results in acceptable limits of the bandwidth utilization and end-to-end delay.

We have also analyzed and presented the capable solution to packet loss problem in OBS networks, while supporting quality-of-service. A new congestion based deflection routing (CBDR) algorithm to reduce the packet loss in OBS networks have been proposed. We evaluated the performance of novel congestion based scheme to reduce burst loss in OBS networks. To reduce the burst loss, contending burst is deflected to another least congested node or in the worst case, back to the previous node after receiving the congestion information from the subsequent node in the path. The proposed technique is tested on variety of performance metrics to find its feasibility as an acceptable solution. Results have shown that the proposed CBDR scheme gives better outcome in case of burst dropping probability and throughput. The proposed scheme avoids contention practically by controlling the burst sending rate according to blockage in the optical network. Consequently the proposed technique is more desirable where better drop ratio while supporting QoS is a vital issue with feasible limits of delay. It has been found that CBDR technique is uncomplicated, practical and its implementation does not lead to any compromise in the basic reason and main inspiration behind the emergence of the OBS archetype, which is simplicity.

To achieve further improvements a novel congestion based scheme, variable size burst routing (VSBR), has been investigated to reduce burst loss in OBS networks. To reduce the burst loss, contending burst is deflected to another least congested node after receiving the congestion information from the subsequent node in the path. Burst dropping is avoided by varying the burst size at the ingress node. The performance of the proposed schemes has also been compared with the existing described schemes. Results have shown that the proposed VSBR scheme gives better outcome in case of burst loss rate but the end-to-end delay is slightly higher as compared to shortest path scheme under consideration. The proposed scheme avoids contention practically by controlling the burst sending rate according to blockage in the optical network. Consequently the proposed technique is more desirable where better burst loss rate performance and blocking probability is a vital issue within acceptable limits of delay.

8.2 RECOMMENDATIONS

The following are some of the possible recommendations based on the individual chapters in the thesis.

The proposed scheduling scheme can be used as a feasible alternative for supporting next generation optical networks, due to high bandwidth utilization. The results achieved have shown that proposed technique uses bandwidth more efficiently as numbers of void checks are enhanced and thus improves the Burst loss ratio and packet loss rate parameters as compared to existing state-of-the-art scheduling algorithms. Better end-to-end performance is another advantage of the proposed scheduling algorithm that makes it suitable for high-performance emerging optical networks.

The proposed burst assembly scheme perform better than the conventional assembly scheme at all congestion levels by varying the threshold limit of the burst according to the traffic conditions in the OBS network. The proposed burst assembly scheme is suitable for transmitting packets which have higher loss and delay constraints. Also, the proposed schemes are easily scalable in order to support multiple classes of traffic. The proposed schemes can therefore be employed in conjunction with all-optical wavelength conversion and buffering through fiber delay lines.

It is observed that the proposed reservations protocols reduce the burstification delay, packet loss, end-to-end delay and at the same time utilize the bandwidth more efficiently. Results show that the proposed schemes are an improved, capable and more efficient solution for OBS networks by improving the bandwidth utilization and providing the most flexible alternative for the next generation optical networks.

Packet/burst loss reduction algorithms to reduce the packet loss have also been analyzed. The proposed techniques have been tested on variety of performance metrics to find its feasibility as acceptable solutions. The proposed schemes avoid contention practically by controlling the burst sending rate according to blockage in the optical network. Consequently the proposed techniques are more desirable where better drop ratio, while supporting QoS, is a vital issue with feasible limits of delay. The proposed schemes when compared to existing schemes provide flexibility during channel reservation based on the type of data to be transmitted and hence are more convenient in performance evaluation as they are easily modifiable according to various network topology and architecture. Thus, the proposed schemes provide flexible solutions suitable for handling the varying traffic demands of the next-generation optical network.

7.3 FUTURE SCOPE

The following are some of the possible areas of future work.

The proposed algorithm uses bandwidth more efficiently as numbers of void checks are enhanced and thus improves the Burst loss ratio and packet loss rate parameters as compared to existing state-of-the-art scheduling algorithms. We believe that it is possible that Burst Drop Ratio can be further improved by assigning bursts of different classes with different lengths and different offset times. We are working on analyzing new techniques and expecting that better performance will be achieved.

We have also considered four different burst scheduling approaches and evaluated their performance for their compatibility with newly introduced Burst Assembly technique. Although, in the OBS literature, burst loss rate is a commonly employed metric, it is observed that the burst loss rates obtained are different than the corresponding packet loss rate. Such a result is caused by variable burst sizes. As a future work, we are focusing on adapting the threshold values of the Burst Assembly algorithm considering the information gathered about the other parts of the network (not only the neighboring links) by using control signals. We believe that such an approach will help a lot under heterogeneous traffic.

Results have shown that the proposed packet loss reduction scheme gives better outcome in case of burst dropping probability and throughput, but the end-to-end delay is slightly higher as compared to conventional scheme under consideration. In such an implementation, it would also be useful to evaluate the desirability of proposed technique where better drop ratio while supporting QoS is a vital issue with feasible limits of delay. The functionality of the scheme can also be extended to provide superior flow and congestion control capabilities in the optical domain.

A signaling framework has been investigated to support entire OBS network services with diverse bandwidth granularities. In addition to the burst loss probability, the end to end delay has been evaluated to implement an improved, capable and more efficient solution for OBS networks. As part of future work the time slots can dynamically be increased/decreased depending on varying traffic conditions.