Parameter inference of queueing models for IT systems using end-to-end measurements
Performance Evaluation
Java Modelling Tools: an Open Source Suite for Queueing Network Modelling andWorkload Analysis
QEST '06 Proceedings of the 3rd international conference on the Quantitative Evaluation of Systems
Estimation for queues from queue length data
Queueing Systems: Theory and Applications
| Hi-index | 0.00 |
Distributed systems with multi-stage workflows are characterized by multiple logical stages which can either execute sequentially or concurrently and a single stage can be executed on one or more physical nodes. Knowing the mapping of logical stages to physical nodes is important to characterize performance and study resource bottlenecks. Often due to the physical magnitude of such systems and complexity of the software, it is difficult to get detailed information about all the system parameters. We show that under light load conditions, the system can be well approximated using first order models and the hence simplifying the system identification problem. For general load, we develop a parameter estimation technique using maximum likelihood and propose a heuristic to solve it efficiently.