This article analyzed advantages and shortages of classical load balancing algorithms based on dynamic feed-back on server cluster, and combined stimulated annealing with this strategy to put forward an optimized mode...This article analyzed advantages and shortages of classical load balancing algorithms based on dynamic feed-back on server cluster, and combined stimulated annealing with this strategy to put forward an optimized model of dynamic load balancing. This model uses stimulated annealing algorithm to calculate accurate performance parameters of load information on every service node, then estimates the actual load of nodes by dynamic feed-back, in order to insure tasks distribution reasonable. Experimental result shows that in the case of large amount of requests, this algorithm, in comparison with classical load balancing strategy of dynamic feedback, can effectively reduce response time of tasks and ensure high throughput which could improve the whole system performance.展开更多
Web serverclusters has become a popular solutionto achieve high performance duetotheirlow price, highreliability,and good scalability. One ofthe key technologiesis “request dispatching”, which is to centrallyaccept...Web serverclusters has become a popular solutionto achieve high performance duetotheirlow price, highreliability,and good scalability. One ofthe key technologiesis “request dispatching”, which is to centrallyacceptallthe incoming HTTPrequests and dispatchthem to the serversinthe clusterto achieve parallelism .The research showsthatthe efficiency ofrequest dispatching is determined by whetherthe serverloads are balanced duringthe dispatching.This paper studies the characteristics ofthe load generated by HTTPrequests,gives anefficient method ofloadsampling,and presentsa novelrequestdispatching algorithm which can predicttheload andbalanceit.The method can be usedto build Web serverclusters ofheterogeneous machines.展开更多
文摘This article analyzed advantages and shortages of classical load balancing algorithms based on dynamic feed-back on server cluster, and combined stimulated annealing with this strategy to put forward an optimized model of dynamic load balancing. This model uses stimulated annealing algorithm to calculate accurate performance parameters of load information on every service node, then estimates the actual load of nodes by dynamic feed-back, in order to insure tasks distribution reasonable. Experimental result shows that in the case of large amount of requests, this algorithm, in comparison with classical load balancing strategy of dynamic feedback, can effectively reduce response time of tasks and ensure high throughput which could improve the whole system performance.
文摘Web serverclusters has become a popular solutionto achieve high performance duetotheirlow price, highreliability,and good scalability. One ofthe key technologiesis “request dispatching”, which is to centrallyacceptallthe incoming HTTPrequests and dispatchthem to the serversinthe clusterto achieve parallelism .The research showsthatthe efficiency ofrequest dispatching is determined by whetherthe serverloads are balanced duringthe dispatching.This paper studies the characteristics ofthe load generated by HTTPrequests,gives anefficient method ofloadsampling,and presentsa novelrequestdispatching algorithm which can predicttheload andbalanceit.The method can be usedto build Web serverclusters ofheterogeneous machines.
