----------------------------------------------------------------------- B O S T O N U N I V E R S I T Y Computer Science Department C O L L O Q U I U M Wednesday, November 21, 11:00 AM (Coffee served at 10:45PM) Seminar Room / MCS 135 Access-Level and POP-Level Traffic Dynamics at a Tier-1 Backbone POP Supratik Bhattacharyya Sprint Labs, Burlingame, CA ABSTRACT : Network traffic measurements provide essential data for networking research and operation. Measurement data is used to study traffic demands, develop models for network traffic, provision networks, and evaluate network performance. While there has been extensive work on network measurement systems, much of it has focused on specialized systems that collect a single type of data, and many of the measurements have been performed on edge networks or in research environments. We have designed a passive measurement system which collects header information from every packet on OC-3, OC-12, and OC-48 network links, and uses GPS synchronized clocks to accurately timestamp each packet. We have collected multiple full day traces at a major POP in Sprint's Tier-1 IP backbone. The first part of the talk will present the design of our monitoring equipment, and briefly describe the ongoing projects at Sprint that utilize this data. The second part of the talk will focus on a particular project that studies traffic dynamics at a POP. Spatial and temporal information about traffic dynamics is central to the design of effective traffic engineering practices for IP backbones. We develop a methodology that combines packet-level traces from access links in the POP and BGP routing information to build components of POP-to-POP traffic matrices. Our analysis shows that there is wide disparity in the geographic distribution of traffic across POPs in the USA. Also, time of day behaviors for different POPs and different access links reveal a high degree of heterogeneity. We study the "elephants and mice" phenomenon at the granularity level of destination address prefixes. We demonstrate that these aggregates exhibit stability throughout the day on per-hour time scales, and thus they form a natural basis for splitting traffic over multiple paths to improve load balancing. Host: Mark Crovella ------------------------------------------------------------------------- For colloquium info, including directions, see http://cs-www.bu.edu/colloquium -------------------------------------------------------------------------