Method and apparatus for load sharing on a wide area network

摘要: Client's ( 106 - 1-106 -N, 107 1-107 -M) on local area networks 102, 103 ) making requests to hot sites, which are connected a wide network 100 such as the Internet, redirected through one of possible plurality different redirectors 101, caching servers (S 1 , S 2 3 ), each have responsibility for mapping or more sites. Each request is probabilistically directed by that map requested site in accordance with weights determined redirector-hot pair so minimize average delay all client across will encounter cached In order determine redirector redirect sites servers, statistics access rates dynamically from traffic flow and reported central management station (CMS) 115 ). Network similarly measured CMS, server computed using queuing model server. Using these parameters inputs, non-linear programming optimization problem solved delay. As rate statistics, well delay, change, algorithm, recalculates forwards them back redirector. other embodiments, redirector-logical item directs may be than identity. For example, logical items can groups clients documents, forwarded web servers.