Cloud computing promises flexibility, high performance and low cost. Despite its prevalence, most datacenters hosting cloud computing services still operate at very low utilization. This is the result of several factors, including interference between applications, platform heterogeneity, and users requesting more resources than they need to avoid performance unpredictability.
A crucial system component that can enable high performance and good system utilization is the cluster manager; the system that orchestrates where applications are placed and how many resources they receive. In this talk, I will describe a new approach in cluster management that relies on two main insights: first, it automates resource management by leveraging practical data mining techniques. Second, it provides a high-level, declarative interface between system and users that centers around performance, not raw resources. Using these insights, I designed and built a datacenter scheduler (Paragon), a cluster manager (Quasar) and scalable provisioning techniques for public clouds. In settings with several hundred servers, I demonstrated that this approach achieves both high application performance and high system utilization. Several production systems, including Twitter and AT&T, have since adopted similar cluster management approaches.
Bio: Christina Delimitrou is a PhD candidate in the EE Department at Stanford University, working in computer architecture and systems. As part of her PhD work, she built practical systems for cluster management and scheduling in large-scale datacenters. She is the recipient of a Facebook Research Fellowship and a Stanford Graduate Fellowship. She has earned an MS from Stanford and a diploma in Electrical and Computer Engineering from the National Technical University of Athens.