A relocation-based heuristic for restoring optical cloud services

Abstract

Optical clouds allow for an integrated management of optical transport and cloud resources (e.g., storage and computing resources running on datacenters). In this paradigm the concept of service relocation (i.e., the ability to re-allocate to a different datacenter node an already provisioned cloud service) offers a new way to restore the network upon failure. This paper presents a heuristic (called H_RELOCATION) based on the service relocation concept to be used for the dynamic restoration of optical cloud services. Upon the occurrence of a network failure, H_RELOCATION solves the routing and resource (i.e., transport plus cloud) allocation problem for each disrupted cloud service allowing, if necessary, to relocate some cloud services to different datacenter nodes. The proposed strategy is benchmarked against two optimal restoration strategies based on Integer Linear Programming (ILP) formulations, the first one without and the second one with the ability to use the service relocation idea. Simulation results show that H_RELOCATION offers performance (i.e., blocking probability, restorability levels and number of required relocations) very close to the optimum, offering on the other hand, a significant reduction in the processing time required for each successfully restored cloud service.