Research On Topology-ware Virtual Network Remapping Algorithm In Cloud Computing Environment

Since the birth of cloud computing, it has a history of nearly10years. Through thedynamic aggregation and reuse of physical resources, Cloud computing greatly improves thephysical resource utilization, and reduces the economic burden of end users. Along thedevelopment of cloud computing, utilization, security and isolation of physical resources andcustomizability of virtual network, etc. draw wide attention in cloud computing center. Howto further improve resource utilization, reduce resource fragmentation and improve theefficiency of network communications has already become a crucial issue.Virtual Network Mapping (VNM) is a core issue to build cloud computing centers, avirtual network (VN) and physical resources reuse. The quality of mapping greatly affects theperformance of VN, such as running efficiency of parallel jobs, resource utilization and loadbalance, QoS, etc.However, most the existing VNM algorithms are based on a fixed topology, and do notconsider the changes of the physical topology. In reality, the substrate topology is likely tochange (physical hosts break down, network cables fail or subnet fails, etc.) in the data center,which may lead to VNM constraints destroying, resulting in a delay or un-completion ofparallel jobs. Therefore, how to maintain VN to adjust to these failures is crucial to bothinfrastructures and service providers.Unfortunately, most existing VNM algorithms cannot be directly applied to maintain VNto endure network failures. On one hand, these algorithms construct an entire VN, which willaffect the healthy virtual subnet and increase the consumption. On the other hand, VNMremapping VN with a fresh start cannot make use of the original mapping, resulting in highercost and worse quality. Therefore, it needs another novel model to enduring the networkfailures. Some algorithms focus on the survivability of VNs for single node failures. Theseworks use a fast recovery mechanism and backup resources for each VN request to ensurenode/link availability. Although they improve the VN survivability, they result in a higherconsumption, because of the lower possibility of single node failure. Thence, a novel model can adaptively discovery the network failures manage the VN,then maintain the invalid subnet according to the original mapping.The paper presents a virtual network remapping adaptive algorithm for evolvingsubstrate topology. It aims to increase the acceptance ratio and remapping similarity betweenthe remapping and original mapping. This model focuses on3points:1. How to increase the remapping quality. A higher similarity means the nearer results ofremapping and original. The higher the similarity is, the better the remapping quality.2. How to reduce the remapping cost. Due to physical network changes resulting inhigher delay of links mapping, how to decrease the hops of physical path is very crucial toimprove the resources utilization.3. How to increase the acceptance ratio. Increasing the acceptance ratio can satisfy moreVN requests to improve the service providers’ benefits.For the above points, the paper proposes a topology-ware VN remapping model based onthe topology-awareness of VN. It makes the following contributions:1. Introduce the concept of Communication Skeleton Network (CSN) to summarize thebackbone topology of a VN. Failed nodes/links are evaluated based on Skeleton Network andremapped in order of importance, which increases the success ratio and improves remappingquality.2. Adopt a physical node evaluation method to select a better candidate node for everyfailed node based on the relationships between invalid and healthy subnets to reduce theresource cost and improve mapping similarity.3. Adopt a preemption idea to further increase the acceptance ratio through preemptingthe healthy subnet resources, while failing in remapping backbone nodes/links.4. Propose a Topology-Aware Virtual Network Remapping algorithm (TA-VNRM) basedon sub-graph isomorphism detection to fulfill the remapping of VN, which makes use of thebacktracking idea.Eventually, the paper simulates the model to verify its effectiveness about the acceptanceratio, remapping quality and the cost.