Capturing the object behaviour for storage system evaluation

Authors:
Chun-Hua Li;Ke Zhou;Dan Feng
Affiliations:
College of Computer Science and Technology, Huazhong University of Science and Technology, Wuhan City, China.;College of Computer Science and Technology, Huazhong University of Science and Technology, Wuhan City, China.;College of Computer Science and Technology, Huazhong University of Science and Technology, Wuhan City, China
Venue:
International Journal of High Performance Computing and Networking
Year:
2010

Citing 8
Cited 0

Measurements of a distributed file system

SOSP '91 Proceedings of the thirteenth ACM symposium on Operating systems principles
Long term distributed file reference tracing: implementation and experience

Software—Practice & Experience
A trace-driven analysis of the UNIX 4.2 BSD file system

Proceedings of the tenth ACM symposium on Operating systems principles
Analysis of Personal Computer Workloads

MASCOTS '99 Proceedings of the 7th International Symposium on Modeling, Analysis and Simulation of Computer and Telecommunication Systems
Passive NFS Tracing of Email and Research Workloads

FAST '03 Proceedings of the 2nd USENIX Conference on File and Storage Technologies
Tracefs: A File System to Trace Them All

FAST '04 Proceedings of the 3rd USENIX Conference on File and Storage Technologies
A comparison of file system workloads

ATEC '00 Proceedings of the annual conference on USENIX Annual Technical Conference
Object-based storage

IEEE Communications Magazine

Quantified Score

Hi-index	0.00

Visualization

Abstract

In this paper, we present a kind of distributed tracing system for evaluating the performance of object storage system. The tracing build on the object-based Lustre file system can capture the object information synchronously from three subsystems of Lustre cluster file system. To improve I/O performance and efficiency, we utilise a queue and a FIFO buffer area to store trace data which is finally written out to a trace file by an output driver. In particular, we use some database tables to save the final integrated traces for convenient post-analysis. We have implemented and evaluated a prototype object tracing system on Linux which patched with Lustre 1.4.5 custom kernel patches. Our evaluation shows a low overhead tracing system.