The Research And Performance Analysis On The Queue Schedule Mechanism Of The Dispatcher In The Auto Self-Allocating Cluster System

With the fast development of network technology and the wide deployment of Internet applications, the network server as the basic platform of Internet information service is confronted with two challenges: Load unceasing increment and Load multiplicity. Cluster system is becoming the main development direction of constructing the high performance network server which relies on the unique superiority in reliability, high performance, extensibility, load balance and cost-effectiveness.National University of Defense Technology brought up a fresh new architecture——Automatic Self-Allocating Cluster System (ASACS), aiming at the flaws which exist in schedule resolutions of the current cluster system based on load balance. Compared with the conventional web cluster, the main characteristic of ASACS is the servers apply for client requests from the dispatcher in a positive manner; replacing traditional load balance and heart beat mechanisms. It concerns with QoS control much more, and provides cluster lever support for the deployment of flexible differentiated service strategies.This paper researches the queue schedule mechanism of dispatcher based on the architecture of ASACS. It proposes a novel stochastic high-level Petri nets (SHLPN) model for the ASACS, and compares with conventional web cluster using the performance analysis technology of SHLPN. Experimental results indicate that ASACS meets the need of QoS well.