LinuxConf Europe logo
LinuxConf Europe 2007
Conference and Tutorials
---------------------------------------------------
Sunday 2nd - Wednesday 5th September
University Arms Hotel, Cambridge, England

LinuxConf Europe 2007

Photos and reports

Timetable

Programme

Registration

LPI Exams

Conference Dinner (Sunday)

Duxford Excursion (Monday)

Exhibitors and Sponsors

Accommodation

Venue

Travel

About Cambridge

Kernel Summit 2007

Other GUUG events

Other UKUUG events

Kimmo Raatikainen - University of Helsinki

A Control Plane Benchmark for Telecommunications Signalling Applications

The world is full of benchmarks. Google provided 2.9 million hits on computer performance benchmark. IEEE Xplore finds 426 articles published since 1.1.2000 in that category. ACM Digital Library has 11,519 entries in that category. Therefore, one can ask whether we still need a new benchmark. Our claim is yes, particularly application specific benchmarks.

Microbenchmarks are useful in pin-pointing performance problems. However, they cannot reliably be used to compare performance of distinct systems in given application domain specific use cases. Well-known microbenchmarks include lmbench for operation system primitives, SPEC CPU2000 for integer (SPECint2000) and floating point (SPECfp2000) calculation, or CommBench, NpBench, and the benchmark suite by the Network Processing Forum for networking. The EDN Embedded Microprocessor Benchmark Consortium (EEMBC) has specified several suites of benchmarks in different application domains to evaluate embedded processors.

Application-level benchmarks exist, but only very few are applicable to benchmark control plane applications in telecom environments. SIPStone is a benchmark for SIP proxy servers. The ETSI TISPAN working group has proposed a telecom benchmark called IMS.

We have developed a new application-level benchmark specification for measuring the performance of different transport, operating system, database, and hardware configurations in control plane applications. In contrast to SIPStone, which measures different SIP proxy server implementations and treats the server as a black box, our benchmark has a fixed server core implementation, but requires the user to provide implementations of the other core components of the server, most importantly the signalling layer and the user location database. This allows us to change only one component at a time and see the difference it has on total system performance in typical control plane usage.

Our benchmark implementation, which is available from Sourceforge, uses two major software packages: OpenSER1 and SIPp2. The implementation has scripts for building, configuring, and running the software from the packages suitably to form a benchmark. The implementation also includes tools for generating and monitoring server load, registering users to OpenSER, analysing the benchmark results, and producing graphs from the results.

The full paper describes the design and implementation in details. We will present our design premises and discuss our development plans.

Submitted paper

Paper (PDF) and Paper (tgz) .


G O L D  S P O N S O R  S I L V E R  S P O N S O R 
Intel
Intel
Google
Google

S  P O N S O R S
Bytemark
Bytemark
Sun
Sun
Novell
Novell
Positive Internet
The Positive Internet Company
collabora
collabora

M  E D I A   S  P O N S O R S
Linux User
Linux User & Developer
Linux Magazine
Linux Magazine
The USENIX Association
The USENIX Association

For more information please contact UKUUG Problems? e-mail webmaster
© Copyright 2007 UKUUG Ltd