Performance Modeling and Anomaly Management for Complex Systems

Dynamic-content online services are comprised of multiple interacting components and data partitions distributed across server clusters. Understanding the performance of these services is crucial for efficient system management and capacity planning. This project investigates profile-driven performance models for multi-component, data-intensive online services. Research has focused on the characterization of inter-component communications as well as concurrent data accesses to storage devices. Built on the performance modeling, research also studied the identification of performance bug manifestations and comprehensive depiction of performance anomaly manifestations over a large space of system configurations and runtime conditions. In particular, our work has resulted in the identification of previously unknown I/O performance bugs in the Linux operating system.

Most Relevant Publications:

Christopher Stewart, Kai Shen, Arun Iyengar, and Jian Yin, "EntomoModel: Understanding and Avoiding Performance Anomaly Manifestations". In Proc. of the 18th IEEE International Symposium on Modeling, Analysis, and Simulation of Computer and Telecommunication Systems (MASCOTS'10), Best Paper Award, Miami Beach, FL, August 2010. Abstract. Postscript version. PDF version.
Kai Shen, "Request Behavior Variations". In Proc. of the 15th International Conference on Architectural Support for Programming Languages and Operating Systems (ASPLOS'10), Pages 103-116, Pittsburgh, PA, March 2010. Abstract. Postscript version. PDF version. Conference talk slides in Powerpoint2007.
Kai Shen, Christopher Stewart, Chuanpeng Li, and Xin Li, "Reference-Driven Performance Anomaly Identification". In Proc. of ACM SIGMETRICS, Pages 85-96, Seattle, WA, June 2009. Abstract. Postscript version. PDF version. Conference talk slides in Powerpoint2007.
Christopher Stewart, Matthew Leventi, and Kai Shen, "Empirical Examination of A Collaborative Web Application". In Proc. of the IEEE International Symposium on Workload Characterization (Special Session on Benchmark Innovation), Seattle, WA, September 2008. Abstract. Postscript version. PDF version. WeBWorK benchmark release.
Christopher Stewart, Terence Kelly, Alex Zhang, and Kai Shen, "A Dollar from 15 Cents: Cross-Platform Management for Internet Services". In Proc. of the USENIX Annual Technical Conference (USENIX'08), Pages 199-212, Boston, MA, June 2008. Abstract. PDF version.
Kai Shen, Ming Zhong, Sandhya Dwarkadas, Chuanpeng Li, Christopher Stewart, and Xiao Zhang, "Hardware Counter Driven On-the-Fly Request Signatures". In Proc. of the 13th International Conference on Architectural Support for Programming Languages and Operating Systems (ASPLOS'08), Pages 189-200, Seattle, WA, March 2008. Abstract. Postscript version. PDF version. Conference talk slides in Powerpoint.
Pin Lu and Kai Shen, "Multi-Layer Event Trace Analysis for Parallel I/O Performance Tuning". In Proc. of the 36th International Conference on Parallel Processing (ICPP'07), XiAn, China, September 2007. Abstract. Postscript version. PDF version.
Chuanpeng Li, Kai Shen, and Athanasios E. Papathanasiou, "Competitive Prefetching for Concurrent Sequential I/O". In Proc. of the Second EuroSys Conference (EuroSys'07), Pages 189-202, Lisbon, Portugal, March 2007. Abstract. Postscript version. PDF version.
Kai Shen, Ming Zhong, and Chuanpeng Li, "I/O System Performance Debugging Using Model-driven Anomaly Characterization". In Proc. of the 4th USENIX Conference on File and Storage Technologies (FAST'05), Pages 309-322, San Francisco, CA, December 2005. Abstract. Postscript version. PDF version. HTML version. Conference talk slides in Powerpoint. Kernel patch code release for bug fix #1 in the paper.
Christopher Stewart and Kai Shen, "Performance Modeling and System Management for Multi-component Online Services". In Proc. of the Second USENIX Symposium on Networked Systems Design and Implementation (NSDI'05), Pages 71-84, Boston MA, May 2005. Abstract. Postscript version. PDF version.

Project Members:

Chuanpeng Li, graduated with Ph.D., now at Bloomberg
Xin Li, University of Rochester
Pin Lu, graduated with M.S., now at Microsoft
Stan Park, University of Rochester
Kai Shen, University of Rochester
Christopher Stewart, graduated with Ph.D., now on the Computer Science faculty at the Ohio State University
Ming Zhong, graduated with Ph.D., now at Google

Contact:

Kai Shen (kshen@cs.rochester.edu)

Support:

This project is supported by the National Science Foundation under Grants No. 0615045 and No. 0448413.

Note:

Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the National Science Foundation.