Research



Major:
Electrical & Computer Engineering

Masters Thesis:
Autonomic utility-based Virtual Machine Management

Duration: May 2008 till date

Thesis advisor:
Dr. Manish Parashar
Associate Professor,
Electrical & Computer Engg Dept,
Rutgers University

Abstract:

Virtual Machine technologies are gaining wide acceptance in modern clusters and data-centers, with efficient ways to dynamically re-configure resources like CPU & memory and migrate OS instances over distinct physical machines. VM migration is considered efficient if it provides a low migration downtime as well as reduced impact on performance of hosted applications. VM resource allocation policies play a vital role in global load balancing over physical machines and optimum use of limited resources.

Currently, most of the VM migration technologies provide Socket Interface and TCP/IP transport to handle migration traffic. Some researchers have proposed VM migration over modern interconnects like InfiniBand, which is slowly changing the server market. Along with providing a high throughput, low latency transport infrastructure for VMs, autonomic management of virtual machines is required to take dynamic decisions about balancing workloads on individual VMs. We are exploring a utility based approach to monitor system performance and trigger corrective action (migration, reconfiguration, etc) if performance constraints are violated.

Work summary till date: