1 A Partial-Protection Approach Using Multipath Provisioning Ananya Das, Charles Martel, Biswanath Mukherjee Department of Computer Science, University of California, Davis, CA 95616 Email: {das, martel, mukherje}@cs.ucdavis.edu Abstract—We study the problem of reliably provisioning traffic using multipath routing in a mesh network. Traditional ap- proaches handled reliability requirements using full-protection schemes. Although full-protection approaches offer high assur- ance, this assurance can be costly. We take a less expensive approach to maintain reliability by offering partial-protection. Specifically, our approach guarantees part of the requested bandwidth, rather than the full amount, in the event of a link failure. We first show that the amount of partial-protection that can be guaranteed is limited by the topology of the network and the bandwidth requirement of a connection request. We then propose an effective multipath algorithm that attempts to provision bandwidth requests while guaranteeing the maximum partial-protection possible. Results show that by effectively se- lecting paths that limit edge overuse, our algorithm achieves very low bandwidth blocking probability. Our algorithm also serves significantly more requested bandwidth than the protection approach. I. I NTRODUCTION Network service providers typically must provision con- nection requests while satisfying Quality-of-Service (QoS) requirements. An important QoS metric is service reliability. Reliability may be measured by how efficiently an application can deal with network failures. Traditional approaches deal with failures using full-protection mechanisms such as path protection [2, 5, 8, 12, 13, 17] and path restoration [3, 6, 10, 11, 13]. Although these approaches offer full recovery in case of network failures, this assurance can be expensive. Instead, network providers can deal with failures by pro- viding part of the requested bandwidth, rather than the full amount, in case a network failure occurs. This approach is referred to as partial-protection. Since partial-protection uses fewer network resources, network providers may offer it to customers who do not require full-protection, and some customers will find partial-protection an attractive option if offered at a lower cost. The term partial-protection has been used to refer to various reliability-aware provisioning schemes [4, 15, 16]. Some of these studies have considered link-level protec- tion [15, 16], however, in this paper we study path-level partial-protection [4]. Specifically, we define partial-protection as a bandwidth provisioning approach that guarantees a spe- cific percentage of the requested bandwidth if a link failure occurs. In the online partial-protection provisioning problem we study, bandwidth requests arrive dynamically and must be scheduled as they arrive. Each request is associated with This work has been supported by NSF Grant No. CNS-0832176. a partial-protection requirement, which is the fraction of bandwidth that must be guaranteed, even in case of failure. For each request, we must provision the requested bandwidth while satisfying the partial-protection guarantee. We consider this problem for high-capacity backbone mesh networks sup- porting virtual concatenation (VCAT) [18]. VCAT enables a connection to be inversely multiplexed on to multiple paths, a feature that has many advantages over conventional single- path provisioning. Multipath provisioning provides better fault tolerance, more effective utilization of network resources, and reduces link congestion. We show that the level of partial-protection possible is limited by the network topology and the bandwidth request. To our knowledge, this is the first work to identify these limits. We then propose an online multipath heuristic to solve this problem. Our algorithm tries to satisfy the maximum amount of partial-protection that can be guaranteed. Unlike previous studies [4, 14], our algorithms guarantee a specific level of partial-protection and can be used to provision connection requests regardless of the network topology and bandwidth requirement. Our algorithm yields low bandwidth blocking probability even at heavy load. In this paper, we study a setting where customers are willing to accept partial-protection, rather than full-protection, in the (rare) case of a network failure. Although full-protection approaches provide complete service even in case of failure, they often require significant network resources. This may result in a decline in overall performance, particularly an increase in bandwidth blocking probability. On the other hand, our algorithm satisfies connection requests while using fewer resources than protection schemes. To understand the cost of full-protection on performance, we compared our algorithms to an efficient full-protection multipath algorithm proposed in [12]. (we describe this heuristic in more detail in Section II). Our results show that our algorithm can provision more of the requested bandwidth than the full-protection scheme. We also find that using a mixture of full and partial protection is more effective than using only full-protection. These results verify that there exists a significant trade-off between full-protection and performance. Therefore, if some customers are willing to accept partial-protection in the (rare) event of a failure, then our approach would be beneficial. The remainder of this paper is organized as follows. Sec- tion II describes some other reliability-aware approaches. Section III describes the provisioning problem that we study and our proposed solutions. Section IV provides our evaluation results. Finally, Section V presents our conclusions.