Types of Load BalancingLoad balancing ensures that each application component is processed by the system best able to handle the component's requirements at that moment, thereby ensuring optimum performance. Load-balancing terminology is new and definitions vary among companies and marketing departments. It is always a good idea to ask specifics when comparing load-balancing products. The definitions in this course are general definitions. Round RobinA popular type of load balancing is called round robin. Round robin indicates that requests are distributed, one per server, until each server is given a request. This distribution is repeated for subsequent requests. With round robin, each server receives an equal amount of requests but there is no evaluation of the servers' resources to handle the requests nor are the requests themselves evaluated. Therefore, a request that could take up to 2 seconds is considered equal to a request that takes only 2 nanoseconds. In the round robin scheme, the assumption is that each request and each server is equal. Weighted Round RobinAnother type of load balancing is called weighted round robin. In a weighted round robin scenario, requests are evaluated and then distributed one request per server. If a request is evaluated to be a more difficult request, the server that receives the request is skipped on the next round of incoming requests. The problem with weighted round robin is that the resources of the server are not evaluated. Therefore, a difficult task might take a server little time if the server has ample CPU and memory resources. In the weighted round robin scheme, the assumption is that the method for evaluating works. NAS Load BalancingIn a NAS load-balancing situation, the servers and their resources are evaluated and then broadcast for comparison with other servers. This means that, as requests go to a server, that server's load changes and thus incoming requests are moved to another server. If a server is not burdened, incoming requests might continue to go to the same, unburdened server. In the NAS load-balancing scheme, the assumption is that the server resources have not changed since the last evaluation and broadcast of the server.
|