Selective hardware/software memory virtualization

Abstract

As virtualization becomes a key technique for supporting cloud computing, much effort has been made to reduce virtualization overhead, so a virtualized system can match its native performance. One major overhead is due to memory or page table virtualization. Conventional virtual machines rely on a shadow mechanism to manage be handled by the guest itself without triggering VM exits. However, the hardware assists do have their disadvantage compared to the conventional shadow mechanism -the page walk yields more memory accesses and thus longer latency. Our experimental results show that neither hardware-assisted paging (HAP) nor shadow paging (SP) can be a definite winner. Despite the fact that in over half of the cases, there is no noticeable gap between the two mechanisms, an up to 34% performance gap exists for a few benchmarks. We propose a dynamic switching mechanism that monitors TLB misses and guest page faults on the fly, and dynamically switches between the two paging modes. Our experiments show that this new mechanism can match and, sometimes, even beat the better performance of HAP and SP.

Revisiting memory management on virtualized environments

“…Many studies in the past have focused on the significant overhead of HAP in 64-bit guests and proposed optimizations from both software and hardware perspectives [Bhargava et al 2008;Barr et al 2010;Wang et al 2011;Ahn et al 2012]. However, our further analysis reveals that HAP performs relatively poorly in 64-bit applications not only because of the larger TLB miss penalty but also because of far fewer VM exits for 64-bit applications using SP compared to the 32-bit version.…”

“…They show that each approach has its own advantage. Wang et al [2011] propose a selective memory virtualization approach that dynamically selects shadow paging and hardware-assisted paging based on VM application behavior. One disadvantage of hardware-assisted paging is the page walk latency to resolve a TLB miss.…”

“…However, virtualization often results in significant performance overhead where memory virtualization overhead plays a key role. Despite continuous efforts to reduce memory virtualization overhead [Adams et al 2006;Menon et al 2006;Santos et al 2008;Zhao et al 2009;Wang et al 2011], it remains significant for certain applications and often the last obstacle among the overall virtualization overhead.…”

“…Surprisingly, with further experiments oriented to explain this phenomenon, we find that during the execution of the 32-bit applications, the major overhead of SP results from VM exits, many of which stem from dynamic memory management in glibc, a C library that is frequently used by C/C++ applications. Although many researchers have already examined the differences between SP and HAP [Wang et al 2011;Ahn et al 2012], they simply ascribe the overhead of SP to hypervisor intervention on page faults. Their studies neither delve further into the root cause of the excessive VM exits nor pay attention to the runtime libraries in guest OSs.…”