2012 Poster Sessions : Hypervisor based DataCenter wide Network Provisioning

Student Name : Vimal Jeyakumar
Advisor : David Mazières
Research Areas: Computer Systems
Abstract:
The last decade saw the rise of cloud computing, where anyone can grab computing and networking power housed in massive data centers. In such multi-tenanted environments, resources such as CPU, memory, disk and network are shared, and it becomes desirable for a cluster-wide operating system to enforce resource isolation policies.

Today, cloud providers allow tenants to provision their CPU, memory and disk (per VM) in measurable units, which is then enforced locally by the hypervisor. For the network, cloud providers have limited means by which a customer can specify network capacity. Dealing with network contention in the cloud is particularly challenging because of (a) the distributed nature of the problem, (b) the large scale of competing entities (1000s of tenants), and (c) the fine timescales at which problems manifest themselves. In this poster, we present a preliminary design of EyeQ—a scalable platform for enforcing networking sharing policies.

Bio:
Vimalkumar is a third year PhD student in Computer Science at Stanford University. His interests are in networks: network protocols, architecture, congestion control; and also network debugging, security and virtualization. In the past, he has interned with the team that built Wolfram|Alpha, and at Microsoft, with Windows Azure's networking team.

Vimalkumar is co-advised by Profs. David Mazières and Balaji Prabhakar. He did his undergrad in Computer Science and Engineering at Indian Institute of Technology Madras.