Neel Shah

Rack Scale Apps

Abstract

Currently, the cloud is one of the most significant trends in computer science and it is projected to become even more important over time. The cloud relies on datacenters which are accountable for 2% of the energy utilization within the US. Sustaining these datacenters involves paying for idle servers that do nothing other than stay powered on. Powering these idle machines costs around $30 billion worldwide.

This is a problem that can effectively be solved by leveraging modern computing architectures such as Rack Scale Architecture. This project is a software version of Rack Scale Architecture that can be used to improve applications’ performance in hardware constrained environments. Rack Scale Apps provides the capabilities to “fool” a process such that it believes it has more memory to work with. Leveraging userfaultfd, a custom memory server, and a network API, Rack Scale Apps makes it possible to pool resources within a rack of servers together and effectively balance the hardware utilization across an entire rack.

Bio:

Neel Shah is a systems hacker interested in operating systems, networking, virtualization, and security. Currently, he is a senior at GWU studying computer science with plans to work at VMware after graduation. At VMware, Neel will work within their CloudOS group.

Presentations

Final Presentation Screencast