Utilizing Memory Content Similarity for Improving the Performance of Replicated Virtual Machines

Abstract

Checkpoint-recovery based Virtual Machine (VM) replication is an emerging approach towards accommodating VM installations with high availability. However, it comes with the price of significant performance degradation of the application executed in the VM due to the large amount of state that needs to be synchronized between the primary and the backup machines. It is therefore critical to find new ways for attaining good performance, and at the same time, maintaining fault tolerant execution. In this paper, we present a novel approach to improve the performance of services deployed over replicated virtual machines by exploiting data similarity within the VM's memory image to reduce the network traffic during synchronization. For identifying similar memory areas, we propose a bit density based hash function, upon which, we build a content addressable hash table. We present a quantitative analysis on the degree of similarity we found in various workloads, and introduce a lightweight compression method, which, compared to existing replication techniques, reduces network traffic by up to 80% and yields a performance improvement over 90% for certain latency sensitive applications. I. INTRODUCTIONWith the recent increase in cloud computing's prevalence, the number of online services deployed over virtualized infrastructures has experienced a tremendous growth. At the same time, however, the latest hardware trend of growing component number in current computing systems renders hardware failures common place rather than exceptional [1]. Replication at the Virtual Machine Monitor (VMM) layer is an attractive technique to ensure fault tolerance in such environments, primarily, because it provides seamless failover for the entire software stack executed inside the Virtual Machine (VM), regardless the application or the underlying operating system. One particular approach, checkpoint-recovery based VM replication, has gained a lot of attention recently [2], [3], [4], [5].Checkpoint-recovery based replication of virtual machines is attained by capturing the entire execution state of the running VM at relatively high frequency in order to propagate changes to the backup machine almost instantly. Essentially, it keeps the backup machine nearly up-to-date with the latest execution state of the primary machine so that the backup can take over the execution in case the primary fails [2].Between checkpoints the VM executes in log-dirty mode, i.e., write accessed pages are recorded so that when the snapshot is taken only pages that were modified in the most

A survey on virtual machine migration and server consolidation frameworks for cloud data centers

“…VM migration rigorously affects the hosted and co-hosted applications' performance due to high resource contention among running applications and VM migration daemon (Hirofuchi et al, 2010). Optimization technologies such as deduplication (Deshpande et al, 2012), compression (Hirofuchi et al, 2010), and delta-based content suppression (Gerofi et al, 2011) work efficiently for specific workload types. These optimization approaches perform poorly when a nonideal workload is hosted.…”

“…Many VM migration approaches have optimized application downtime (Gerofi et al, 2011;Hines et al, 2009;Svärd et al, 2011) and total migration duration (Deshpande et al, 2012, Gerofi et al, 2011Hines et al, 2009;Koto et al, 2012;Svärd et al, 2011;Zhang et al, 2013) by employing optimization and avoiding aggressive migration termination (a case of pre-copy). However, only few schemes (Gerofi et al, 2011, Svärd et al, 2011 consider the application's QoS degradation while migrating VMs across servers within a DC. (Yin et al, 2014) N Table 3 illustrates the objective of VM migration schemes.…”

“…For instance in (Deshpande et al, 2012) entire rack migration was considered while provisioning maintenance at the rack level. Similarly, in (Gerofi et al, 2011), replicating a VM across cloud servers was considered to avoid service unavailability. In (Sahni and Varma, 2012), mixed workload hosted inside VM was taken into account while migrating VMs from one DC to another.…”

“…However, since VM migration is a resource-intensive process, application performance is significantly affected during migration (Barhamet al, 2003;Clark et al, 2005;Xu et al, 2014). Bandwidth optimization techniques such as deduplication, compression, write-throttling, and dynamic rate-limiting optimize bandwidth utilization efficiency to enhance application performance (Deshpande et al, 2012;Gerofi et al, 2011, Hirofuchi et al, 2010Sahni and Varma, 2012;Svärd et al, 2011). Furthermore, VM migration techniques migrate VMs either within LAN or WAN boundaries.…”

Tradeoffs in compressing virtual machine checkpoints

“…Note that similarity compression examines the similarity in the memory of live, executing VMs continuously, different from VMFlock [9], which utilizes the similarity in VM disks during migration of static VM images. Instead of finding "identical" data in dirty pages, Gerofi et al find memory areas that are "similar" to the dirty pages, and send only the differences between the dirty pages and these memory areas to reduce checkpoint replication traffic [15]; their approach is currently applied to each VM independently. The same idea has also been used to reduce VM migration traffic [32,14].…”