Simple generational garbage collection and fast allocation
Software—Practice & Experience
Concurrent Programming in Java: Design Principles and Patterns
Concurrent Programming in Java: Design Principles and Patterns
Computer
Thread Transparency in Information Flow Middleware
Middleware '01 Proceedings of the IFIP/ACM International Conference on Distributed Systems Platforms Heidelberg
Initial Observations of the Simultaneous Multithreading Pentium 4 Processor
Proceedings of the 12th International Conference on Parallel Architectures and Compilation Techniques
Power Efficient Processor Architecture and The Cell Processor
HPCA '05 Proceedings of the 11th International Symposium on High-Performance Computer Architecture
A study of Java object demographics
Proceedings of the 7th international symposium on Memory management
Performance Studies of Commercial Workloads on a Multi-core System
IISWC '07 Proceedings of the 2007 IEEE 10th International Symposium on Workload Characterization
Addressing Cache/Memory Overheads in Enterprise Java CMP Servers
IISWC '07 Proceedings of the 2007 IEEE 10th International Symposium on Workload Characterization
| Hi-index | 0.00 |
Multi-core platforms are becoming prevailing in telecom infrastructures, and many SIP(Session Initiation Protocol) enabled applications are using Java as the development language and runtime environment. It is important to understand the workload characteristics and performance issues of Java based SIP stack over multi-core platforms. In this paper, two RFC 3261 compliant Java based SIP stacks, the JAIN-SIP and a proprietary SIP stack are studied in depth. Especially we focused on the typical performance issues caused by either Java language features or the workload of SIP protocol, including scalability of SIP stack, resource contention issue and GC impact with SIP memory usage pattern. Some optimization techniques and their drawbacks are also discussed along with the performance evaluation. It turns out that due to the complex combination of SIP protocol semantics and Java language features, the performance of Java based SIP stack tends to be far from competitive on multicore platforms. Extensive optimization and certain improvements of program structure and object management policy may help, however, may, on the other hand, sacrifice some key values of the java language, e.g. easy development.