Broadband services today demand outstanding performance of the broadband networks to carry huge traffic and, at the same time, to detect an interruption due to network element failures and to survive from that critical condition. In a Wavelength Division Multiplexing (WDM) network with data rate of 100 Gbps (or even nearly 50 Tbps), the impact on the link or node failure is exacerbated by the extremely-high volume of traffic. An interruption means a huge revenue loss. Thus, optical link survivability has an imperative role in the failure handling in high-speed optical networks.
The optical network restoration is categorized into three types: link, path and subpath restoration. In optical network survivability, link-based and subpath-based restorations have been discussed in many papers. One problem of all those restorations is that a connection would be dropped due to the unavailable network resources.
This thesis proposes Link-Based and Subpath-Based Hybrid Restoration scheme (LINKSPATH). It is a hybrid restoration scheme formed by two well-known restoration methods, the link-based and subpath-based methods. Both restoration schemes utilize Dijkstra algorithm in calculating shortest route (of link-based) and shortest subpath (of subpath-based) and are run in parallel. All alternative routes and subpath are pre-calculated in connection setup session. LINKSPATH also calculates recovery time for both alternative route and subpath. The calculation results of these two methods are compared and stored in a buffer as two options with different priority. In case of a link failure occurs, the connection is rerouted to the previously calculated backup route or subpath.
LINKSPATH applies a selection process to choose a pre-calculated backup route with the smallest recovery time and tag it as primary backup and secondary backup for the other one. In case of the primary backup is unavailable the secondary backup then takes the role. In worst case, the second backup route or path might be unavailable as well, due to another link failure or node device failures. In this condition the traffic would be dropped and retransmission request would be sent to the source node. In this paper, the retransmission is assumed to be an end-to-end path-based restoration and would be compared to LINKSPATH to see the significance of this hybrid scheme.
In the experiment, 16 nodes and 22 bidirectional links NSFNET network model was used. The experiment shows that the LINKSPATH scheme has maximum restoration time of 122,40 ms and minimum value of 23.36 ms with the average restoration time of 72.27 milliseconds. The maximum delay due to LINKSPATH restoration is 148.87 ms, which is below 200 milliseconds of ANSI TR1.TR.68-2001 standard. This indicates that LINKSPATH would not jeopardy communication.
Most of LINKSPATH restorations have single alternative route, rather than two routes. Very small part of restorations with two alternative routes have full-backup redundancy. This means the full redundancy for all restorations as main objective could not be achieved.
Most of restorations utilized subpath-based rather than link-based. It is due to backhauling of link-based method (in some cases) and the system default which takes subpath-based when both methods result the same restoration time. However, both methods have had a role as backups to each other.
Compared to link-based and subpath-based individual performance, LINKSPATH achieved larger number of restorations with significance over retransmission time more than twofold and threefold. It is proven that LINKSPATH type of hybrid scheme has better performance than the forming methods.
The ratio of retransmission to LINKSPATH restoration time has to be more than one. If not, then it would reduce the effectiveness of the sceheme. From the experiment, LINKSPATH effectiveness is only 78.62%. This means 21.38% of all restorations have larger restoration time compared to retransmission.