COWCache: effective flash caching for Copy-on-Write virtual disks

Abstract

Host-side flash caching emerges as an effective solution for improving the performance of virtual machines (VMs) in cloud computing environment. However, for VMs with the commonly used Copy-on-Write (COW) virtual disks, flash caching in fact has negative impacts since it brings lots of unnecessary cache writes, hurting both the VM performance and the flash endurance. This paper presents COWCache, a COW-aware caching solution that addresses this problem by co-designing flash caching with COW-based virtual disks. First, COWCache designs a new architecture that bridges the semantic gap between flash caching and virtual disk management for realizing the VMs' real data locality. Second, it separately manages COW metadata with fine-grained caching and journaling to improve the metadata caching efficiency. Third, it provides a novel decoupled Copy-on-Write mechanism, which decouples the amplified data requests from the critical I/O path and only admits the data with real VM locality into flash cache. COWCache also introduces a new data structure, the virtual cache map, to mitigate the memory footprint overhead for indexing the cached data in flash. Evaluations show that COWCache improves the application performance by up to 122.7% and reduces the flash cache writes by up to 78.5% compared to traditional flash caching solutions.

JCache: Journaling-Aware Flash Caching

“…Flash-based SSDs are being increasingly deployed in cloud computing environments as shared flash caches for concurrently running VMs to accelerate their storage performance [1], [4], [8], [10], [12], [13], [15], [17], [20], [21], [24]. Due to the easy control for the VMs and sharing the flash caches, managing flash caches outside of the VMs (e.g., in the hypervisor) is a commonplace, as discussed in [4], [8]. As illustrated in Figure 1, after deploying flash caching for VMs, the cache manager receives all the I/O requests issued from the VMs, sends them to the virtual disk backend when necessary, and caches them with various replacement policies.…”

“…In this section, we discuss how JCache recovers the file system in the VMs to a consistent state in case of guest OS or host failures. Although there are other failure recovery issues related to flash caching as introduced in [8], [12], [24], they are out of the scope of this paper, and here we mainly concentrate on the journaling part. The general recovery procedure for the file system in the guest OS does not need any changes, and JCache can correctly process all the journal I/O requests sent by the VMs.…”

“…CacheDedup [15] and [7] employ cache deduplication to improve the VMs' I/O performance. COWCache [8] co-designs flash caching with the management of Copy-on-Write virtual disks to improve the VMs' caching performance. In [12], [24], the authors optimize the cache writeback policies for flash caching.…”

“…While deploying flash caching for VMs that mostly use journaling file systems, we observe a three-fold challenge to fully exert the high performance and persistence properties of flash caches. First, the cache manager that manages the shared flash caches for VMs is usually deployed outside of the VMs for simplicity and flexibility (e.g., in the hypervisor [3], [4], [8], [13], [20]), such that the semantics of journaling in the VMs' file system are transparent to flash caching, which makes it difficult to optimize the VM's journaling in the caching layer. Second, the logging I/Os are just considered as regular I/O requests by the cache manager, which then will write these I/O requests to the VMs' virtual disk backend stored in the shared storage server and cache them in flash as well.…”