Publication | Closed Access
An analysis of Linux scalability to many cores
342
Citations
31
References
2010
Year
Cluster ComputingEngineeringLinux ScalabilityComputer ArchitectureSoftware EngineeringSoftware AnalysisSloppy CountersParallel ComputingComputer EngineeringComputer ScienceScalability ReasonPerformance Analysis ToolScalable ComputingPerformance ScalabilityScalability BottlenecksOperating SystemsProgram AnalysisParallel Performance EvaluationCloud ComputingMany-core ArchitectureMultiprocessor SystemParallel ProgrammingPerformance PortabilitySystem Software
The study examines scalability of seven Linux applications on a 48‑core machine and shows that standard parallel techniques plus a new sloppy‑counter method can eliminate kernel bottlenecks or be avoided by slight application changes. The authors evaluate scalability on the 48‑core system, applying standard parallel programming methods and the sloppy‑counter technique to remove kernel bottlenecks or shift them to application modifications. All but gmake trigger kernel bottlenecks, fixing them required 3002 lines of code, yet the analysis suggests traditional OS designs remain scalable.
This paper analyzes the scalability of seven system applications (Exim, memcached, Apache, PostgreSQL, gmake, Psearchy, and MapReduce) running on Linux on a 48- core computer. Except for gmake, all applications trigger scalability bottlenecks inside a recent Linux kernel. Using mostly standard parallel programming techniques-- this paper introduces one new technique, sloppy counters-- these bottlenecks can be removed from the kernel or avoided by changing the applications slightly. Modifying the kernel required in total 3002 lines of code changes. A speculative conclusion from this analysis is that there is no scalability reason to give up on traditional operating system organizations just yet.
| Year | Citations | |
|---|---|---|
Page 1
Page 1