March 28, 2012
Title: System Support for Agile Cloud Computing
Abstract: While cloud computing has taken the information technology landscape by storm, the painful reality is that developing applications that take full advantage of the flexibility promised by the cloud model is still very challenging. Whereas the clouds pay-as-you-go model makes it possible to develop servers that dynamically resize their resource allocation, existing mechanisms lack the agility to let server grow and shrink in tight concert with user demand. Foremost, creation of new virtual machines is slow new servers take a while to boot because its a laborious I/O bound task. Moreover, this latency is hard to predict
instantiation latencies in Amazons EC2 cloud have been observed to fluctuate sharply around a two-minute mean. Furthermore, once booted, the servers performance-critical application and OS ca! ches are essentially empty, which degrades performance when it is most needed to service demand spikes. Therefore, server owners have incentives to keep VMs active for long periods, both to provide slack resources during long instantiation latencies, and
because servers with large, warm buffers become crucial to overall performance and are too valuable to sacrifice.
This talk presents research on cloud primitives that make it possible for users to fully leverage the promise of the cloud while reducing the effort required for developing and deploying cloud applications. I will describe VM fork, a new abstraction that can replicate a VM into hundreds of cloud hosts in less than
a second, and will show how it can be leveraged to run a wide range of cloud workloads including high performance applications, web servers and databases.
Biography: Eyal de Lara is an Associate Professor in the Department of Computer Science at the University of Toronto. Eyal received his PhD and MSc
from Rice University in 2002 and 1999, and a BSc from the Instituto Tecnologico de Monterrey in 1995. His research interests include distributed systems and mobile computing. His research has been recognized with an IBM Faculty Award and a NSERC Discovery Accelerator Award.