Nedeljko Vasic received his Ph.D. from the School of Computer Sciences and Communications, from EPFL, Switzerland, and his MSc degree in school of Computer Science and Automatics from University of Novi Sad (FTN), Serbia. In 2010, he held a position at IBM Research, Zurich. Since May 2011, he has been working as a post-doctoral researcher at the Networked Systems Laboratory, and Operating Systems Laboratory, EPFL. His main interests are in: i) performance evaluation and resource management in virtualized environments, and ii) Internet and data center architectures that result in better performance, energy-efficiency, and elasticity. Nedeljko is a recipient of: i) a prestigious award for the best graduated student of University of Novi Sad (Republic of Serbia), ii) the Best Paper Award at COMSNETS 2009, iii) an IBM PhD Fellowship in 2010, and iv) an Honorable Mention in the 2012 EuroSys Roger Needham PhD Award competition for the best systems PhD in Europe.

AG 1, AG 2, AG 3, AG 4, AG 5, SWS, RG1, MMCI   Info

Expert Audience

English

60 Minutes

Saarbrücken

Cloud computing in general and Infrastructure-as-a-Service (IaaS) in particular, are becoming ever more popular. However, effective resource management of virtualized resource is a challenging task. Moreover, performance interference (and the resulting unpredictability in the delivered performance) across virtual machines co-located on the same physical machine threatens to make cloud computing inadequate for performance-sensitive customers and more expensive than necessary for all customer.

In this talk, I will describe two frameworks - DejaVu and DeepDive - for dealing with the resource management and performance interference issues in virtualized environments. The key idea behind DejaVu is to cache and reuse the results of previous resource allocation decisions at runtime. By doing so, it speeds up adaptation to workload changes by 18X relative to the state-of-the-art. DeepDive transparently diagnoses and manages performance interference in the cloud by leveraging easily-obtainable low level metrics to discern when interference is occurring and what resource is causing it. DeepDive also mitigates interference using a low-overhead approach to identifying a VM placement that alleviates interference.

Brigitta Hansen

0681 93039102

--email hidden

Yes

Kaiserslautern

G26

112

passcode not visible

logged in users only