2022 · 2022 IEEE International Conference on Cloud Engineering (IC2E)
Poster Paper: Operating System Support for Applications Performance Analysis
Evidence basis: full-text-reviewed · Review status: catalog-reviewed; paper-author approval pending
performance-analysis system-tracing kernel-tracing
operating-system support application performance analysis performance instrumentation IC2E 2022 poster paper
Core contribution: This poster proposes operating-system-aware antipattern detection that uses system-level traces and scalability trends to expose resource blocking that aggregate CPU utilization can hide.
Problem and motivation
Existing performance-antipattern detectors may not see OS-wide causes. A workload can show acceptable aggregate CPU utilization while requests wait on a database, mutex, disk, or another resource, so exact critical-path evidence is needed.
Method and contribution
The poster studies the One-Lane Bridge and extends the idea to an N-Lane Bridge. Starting with one request, the method increases concurrency until response time reaches ten times the baseline, examines the response-time trend, and uses the critical path to identify blocking events. The case study uses an open-source Flask/MySQL CRUD application and HTTP add-user requests. Stress tests include sysbench CPU prime-to-20000 and 150 GB random-read/write I/O for 60 seconds, with system-level tracing used to quantify the surrounding behavior.
Findings and evidence
At approximately 500 concurrent requests, response time reaches the stated 10x boundary; the critical path is dominated by a recv system call while the target thread waits for the backend database, supporting an OLB diagnosis linked to database concurrency limits. The poster reports tracing overhead of 0.0098% for the CPU case, 7.852% for read I/O throughput, and 7.829% for write I/O throughput.
Limitations and future directions
Limitations: This is a two-page poster/short paper with one application and limited validation. The heuristic boundary and antipattern interpretation are case-specific, and the evidence does not establish detection accuracy across a broad set of systems or resource configurations.
Future work: Implement and evaluate Traffic Jam and Ramp antipattern detectors, test more applications and database/resource configurations, and report ground-truth agreement, detection delay, and false-positive behavior.
Resources
Sources and identifiers
- Published version published
- DBLP record public_source_record
When to cite this paper
Cite this poster when motivating OS-aware One-Lane/N-Lane Bridge performance-antipattern detection.
- The response-time scaling rule and critical-path diagnosis.
- The Flask/MySQL concurrency case and recv/database wait evidence.
- The CPU and I/O tracing-overhead measurements.
- A short-paper case study rather than broad detector validation.
Citation
@inproceedings{ezzatiJivan2022posterpaper,
author = {Riley VanDonge and Naser Ezzati-Jivan},
title = {Poster Paper: Operating System Support for Applications Performance Analysis},
year = {2022},
booktitle = {2022 IEEE International Conference on Cloud Engineering (IC2E)},
pages = {279-280},
publisher = {IEEE},
doi = {10.1109/IC2E55432.2022.00039},
url = {https://doi.org/10.1109/IC2E55432.2022.00039}
}Other citation formats for Word and reference managers
VanDonge, R., & Ezzati-Jivan, N. (2022). Poster Paper: Operating System Support for Applications Performance Analysis. In 2022 IEEE International Conference on Cloud Engineering (IC2E) (pp. 279-280). https://doi.org/10.1109/IC2E55432.2022.00039R. VanDonge and N. Ezzati-Jivan, "Poster Paper: Operating System Support for Applications Performance Analysis," in 2022 IEEE International Conference on Cloud Engineering (IC2E), pp. 279-280, 2022, doi: 10.1109/IC2E55432.2022.00039