End-to-End Diagnosis of Performance Problems
in Enterprise Networks
by Victor BAHL

Abstract

Enterprise network management is a billion dollar industry and localizing the sources of performance problems in large enterprise networks remains a challenging problem. Existing management systems either manage applications running on servers or manage the network elements. The arduous task of correlating the network and application level behavior is left to the network operators. 

We take an end-to-end approach to network management.  Software running on end hosts extracts network dependency graphs of applications and services running on them. We then use these graphs to correlate the behavior of distributed application, services and core network elements for fast, accurate problem localization. Dependencies are numerous, complex and inherently multi-level, spanning hardware and software components across the network and the computing infrastructure.  For faster analysis, we introduce the idea of Inference Graphs, which captures user-perceptible problems rooted in conditions that give rise to both partial service degradation and hard faults.  These techniques are a part of our PMON Enterprise Network Management System. I will describe the PMON project and show how we are using it to diagnose hard to localize performance problems in over 3100 line of business applications.

About the speaker

Victor BAHL is a Principal Researcher and Founding Manager of the Networking Research Group in Microsoft Research. His personal research interests span a variety of topics in wireless systems design, mobile networking, and network management. He has built and deployed several highly cited networked systems and has authored over 80 papers and 120 patent applications (of which 52 have issued).  He is the founder and past-Chair of ACM SIGMOBILE, the founder and past Editor-in-Chief of ACM Mobile Computing and Communications Review and the founder and steering committee chair of the Mobile Systems Conference. He has served as a General Chair of SIGCOMM and MobiCom and is serving on the steering committees of numerous other IEEE & ACM conferences.  He is on the board of over half-a dozen journals,  on several NSF and NRC panels, and on over six dozen program committees. Dr. Bahl received Digital's Doctoral Engineering Award in 1995 and SIGMOBILE's Distinguished Service Award in 2001. In 2004, Microsoft nominated him for the innovator of the year award. He is an ACM Fellow and an IEEE Fellow. More on him at http://research.microsoft.com/~bahl/.