Optimizing Cost and Performance in Online Service Provider Networks

Thumbnail

Event details

Date 08.04.2011
Hour 10:30
Speaker Prof. Y. Charlie Hu, Purdue University
Location
Category Conferences - Seminars
To satisfy global demand, an online services providers (OSPs) such as Google operates a network of geographically dispersed data centers that connects with many Internet service providers (ISPs). Two key considerations for the OSP are the cost and the performance of delivering traffic to its millions of users. We present the first traffic engineering system for OSPs which jointly optimizes the cost and the performance of delivering traffic from an OSP network to its users. Our system is based on two key techniques. First, it uses a novel route-injection mechanism to measure the performance of alternative paths that are not being currently used, without disturbing current traffic. Second, based on the real-time cost, performance, traffic, and link capacity information, it computes the optimal cost vs. performance curve for the OSP where each point on the curve represents a potential operating point for the OSP. The OSP can then pick the operating point that represents the desired cost-performance trade-off. Our trace-driven evaluation in a large OSP with 11 data centers shows that our system can reduce the OSP's traffic cost by 40% without any increase in path latency. This is a joint work with Microsoft Research.

Practical information

  • General public
  • Free

Contact

  • Madeleine Robert

Event broadcasted in

Share