rtpengine集群_sip服务器集群系统的负载均衡技术分析-analysis of load balancing technology for sip server cluster system....

sip服务器集群系统的负载均衡技术分析-analysis of load balancing technology for sip server cluster system

AbstractWith the more and more drastic competition in the global communication market, no matter what you are, the traditional Internet vendors or the fast growing mobile Internet vendors, you have to switch your role of a platform or pipe provider over to the role of service provider to stand out. It’s of vital importance to provide high quality service to enhance user stickiness, whereas building a high availability service platform is one of the most critical step. The mobile Internet service are controlled mainly by SIP, bringing out higher demand for real-time feature. Thus, there exists many more challenges in system throughput and redundancy for the SIP server cluster system comparing with other cluster system such as web cluster, database cluster and file system cluster.After researching on the existing commercial and open source project about the SIP server cluster system, we know a lot about the detailed technology used in the system and also find that there exists much room for improvement. Focusing on the key technique used in the SIP server cluster, a new load balancing algorithm is presented based on the adequate work of load balancing research. Also, we evaluate several load balancing algorithm theoretically. To verify the performance of our new algorithm, we present the design, implementation and evaluation of our system using a cluster of Intel x86 machines running Linux. The main work includes the design and implementation of SIP load balancer and SIP server. The detailed test cases are presented for function test and performance test.The load balancer in our system can distribute the SIP requests to the SIP servers evenly. When one of the load balancer crashes, the other backup server will take over seamlessly. Besides, the system is adapted to the SIP server’s dynamic change. The new algorithm, least work left based on SIP transaction, achieves its performance by dynamically estimating the back-end SIP server load precisely.Keywords：SIP, Cluster,