Joint Tracking of Performance Model Parameters and System Behavior Using a Multiple-Model Kalman Filter

Online resource management of a software system can take advantage of a performance model to predict the effect of proposed changes. However, the prediction accuracy may degrade if the performance model does not adapt to the changes in the system. This work considers the problem of using Kalman filters to track changes in both performance model parameters and system behavior. We propose a method based on the multiple-model Kalman filter. The method runs a set of Kalman filters, each of which models different system behavior, and adaptively fuses the output of those filters for overall estimates. We conducted case studies to demonstrate how to use the method to track changes in various system behaviors: performance modeling, process modeling, and measurement noise. The experiments show that the method can detect changes in system behavior promptly and significantly improve the tracking and prediction accuracy over the single-model Kalman filter. The influence of model design parameters and mode-model mismatch is evaluated. The results support the usefulness of the multiple-model Kalman filter for tracking performance model parameters in systems with time-varying behavior. key words: performance modeling, tracking filter, multiple-model method, queueing theory, resource management

[1]  Marin Litoiu,et al.  The use of optimal filters to track parameters of performance models , 2005, Second International Conference on the Quantitative Evaluation of Systems (QEST'05).

[2]  Thia Kirubarajan,et al.  Estimation with Applications to Tracking and Navigation: Theory, Algorithms and Software , 2001 .

[3]  Martin F. Arlitt,et al.  Web server workload characterization: the search for invariants , 1996, SIGMETRICS '96.

[4]  Marin Litoiu,et al.  Performance Model Estimation and Tracking Using Optimal Filters , 2008, IEEE Transactions on Software Engineering.

[5]  Antonio M. Pascoal,et al.  Issues, progress and new results in robust adaptive control , 2006 .

[6]  Asser N. Tantawi,et al.  Estimating Model Parameters of Adaptive Software Systems in Real-Time , 2010 .

[7]  Inseok Hwang,et al.  Observability criteria and estimator design for stochastic linear hybrid systems , 2003, 2003 European Control Conference (ECC).

[8]  Eduard Ayguadé Parra,et al.  A Survey on Performance Management for Internet Applications , 2010 .

[9]  Marin Litoiu,et al.  Tracking time-varying parameters in software systems with extended Kalman filters , 2015, CASCON.

[10]  Marin Litoiu,et al.  Integrated estimation and tracking of performance model parameters with autoregressive trends , 2011, ICPE '11.

[11]  R. Lynn Kirlin,et al.  Robust Adaptive Kalman Filtering with Unknown Inputs , 1986, 1986 American Control Conference.

[12]  Asser N. Tantawi,et al.  Enhanced inferencing: estimation of a workload dependent performance model , 2009, VALUETOOLS.

[13]  D. Simon.  Optimal State Estimation: Kalman, H Infinity, and Nonlinear Approaches , 2006 .

[14]  Hui Zou,et al.  Combining time series models for forecasting , 2004, International Journal of Forecasting.

[15]  Asser N. Tantawi,et al.  An analytical model for multi-tier internet services and its applications , 2005, SIGMETRICS '05.

[16]  Eduard Ayguadé,et al.  A survey on performance management for internet applications , 2010 .

[17]  Gregory R. Ganger,et al.  Ironmodel: robust performance models in the wild , 2008, SIGMETRICS '08.

[18]  Evgenia Smirni,et al.  MAP-AMVA: Approximate mean value analysis of bursty systems , 2009, 2009 IEEE/IFIP International Conference on Dependable Systems & Networks.

[19]  Asser N. Tantawi,et al.  Real-time performance modeling for adaptive software systems , 2009, VALUETOOLS.

[20]  Jens Happe,et al.  Performance evaluation of scheduling policies in symmetric multiprocessing environments , 2009, 2009 IEEE International Symposium on Modeling, Analysis & Simulation of Computer and Telecommunication Systems.

[21]  Asser N. Tantawi,et al.  CPU demand for web serving: Measurement analysis and dynamic estimation , 2008, Perform. Evaluation.

[22]  Marin Litoiu,et al.  Service System Resource Management Based on a Tracked Layered Performance Model , 2006, 2006 IEEE International Conference on Autonomic Computing.

[23]  V. Jilkov,et al.  Survey of maneuvering target tracking. Part V. Multiple-model methods , 2005, IEEE Transactions on Aerospace and Electronic Systems.

[24]  X. Rong Li,et al.  A comparative study of multiple-model algorithms for maneuvering target tracking , 2005, SPIE Defense + Commercial Sensing.

[25]  K. Do,et al.  Efficient and Adaptive Estimation for Semiparametric Models. , 1994 .

[26]  Dan Simon,et al.  Optimal State Estimation: Kalman, H∞, and Nonlinear Approaches , 2006 .

[27]  Richard Wolski,et al.  The network weather service: a distributed resource performance forecasting service for metacomputing , 1999, Future Gener. Comput. Syst..

[28]  Y. Bar-Shalom,et al.  The interacting multiple model algorithm for systems with Markovian switching coefficients , 1988 .

[29]  Paola Inverardi,et al.  Model-based performance prediction in software development: a survey , 2004, IEEE Transactions on Software Engineering.