The Department of Computer Science at the University of Cyprus cordially invites you to the Colloquium entitled:

Efficient Inter-datacenter bulk transfers or How to Book Some Terabytes on 'Red-Eye' Bandwidth


Speaker: Dr. Michalis Sirivianos
Affiliation: Telefonica Research Barcelona, Spain
Category: Colloquium
Location: Room 148, Faculty of Pure and Applied Sciences (FST-01), 1 University Avenue, 2109 Nicosia, Cyprus (directions)
Date: Thursday, February 3, 2011
Time: 12:00-13:00 EET
Host: George Pallis (gpallis AT

Large datacenter operators with sites at multiple locations dimension their key resources according to the peak demand of the geographic area that each site covers. Notably, the demand of each area is known to follow strong diurnal patterns with high peak to valley ratios that result in poor average utilization across a day. In this paper, we show how to rescue transit bandwidth across multiple datacenters by using it for non-real-time applications, such as backups, propagation of bulky updates, and migration of data that improve fault tolerance, end-user experience, and energy/personnel costs, respectively. Achieving the above is non-trivial since leftover transit bandwidth appears at different times, for different durations, and at different places in the world. For this purpose we have designed, implemented, and validated a system that employs a network of storage nodes to stitch together unutilized bandwidth, whenever and wherever it exists. Our system employs advanced store-and-forward algorithms that schedule data transfers across time and space. It uses information on the availability of leftover resources and is able to react to fluctuations and failures. We have extensively compared our system with other bulk transfer mechanisms such as direct transfer, multipath forwarding, and naive store-and-forward using both emulation and a live deployment in a real CDN. Our evaluation shows that \hermes\ outperforms all other mechanisms and can rescue up to x5 additional datacenter bandwidth thus making it a valuable tool for datacenter providers.

Short Bio:
Michael Sirivianos is a Jr. Researcher at Telefonica Research, Barcelona. He earned a PhD in Computer Science from Duke University in 2010. He received a B.S in Electrical and Computer Engineering from the National Technical University of Athens in 2002, and an M.S. in Computer Science from the University of California, San Diego in 2004. His research interests include introducing social trust in distributed system design, cooperative content distribution and human verifiable secure device pairing.

  Mailing List:

Sponsor: The CS Colloquium Series is supported by a generous donation from Microsoft