Memory Resource Balancing For Virtual Machines On One Host

Virtualization has become a common abstraction layer in modern data centers. Bymulti-plexing hardware resources into multiple virtual machines (VMs) and thus enablingseveral operating systems to run on the same physical platform simultaneously, it caneffectively reduce power consumption and building size or improve security by isolatingVMs.In a virtualized system, memory resource management plays a critical role in achievinghigh resource utilization and performance. Insufficient memory allocation to a VM willdegrade its performance dramatically. On the contrary, over-allocation causes waste ofmemory resources. Meanwhile, a VM’s memory demand may vary significantly. As a result,effective memory resource management calls for a dynamic memory balancer,which,ideally, can adjust memory allocation in a timely manner for each VM based on theircurrent memory demand and thus achieve the best memory utilization and the optimaloverall performance.In order to estimate the memory demand of each VM and to arbitrate possible memoryresource contention, a widely proposed approach is to construct an LRU-based miss ratiocurve (MRC), which provides not only the current working set size (WSS) but also thecorrelation between performance and the target memory allocation size. In this dissertation,we first present a low overhead LRU-based memory demand tracking scheme, whichincludes three orthogonal optimizations: AVL-based LRU organization, dynamic hot setsizing and intermittent memory tracking. Our evaluation results show that, for the wholeSPEC CPU2006benchmark suite,after applying the three optimizing techniques, the meanoverhead of MRC construction is lowered from173%to only2%.Based on the measured current WSS, we then predict its trend in the near future andtake different memory allocation scheduling policies for different prediction results, so theamount of physical memory on the local host for maximum resource utilization.