O. Kremien, J. Kramer, and J.N. Magee. Scalable and Adaptive Load Sharing for Distributed Systems. IEEE Parallel and Distributed Technology, 1(3), 1993.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....will point back to that site. Therefore, the service for this client has effectively been moved to the preferred site. 4 Local Load Sharing The purpose of load sharing is to reduce the average response time by smoothing out transient peak overload on some of the nodes of a distributed system [4]. The WOMsprayer is a software TCP connection router, that serves as a load sharing frontend to a collection of back end servers. The WOMsprayer host offers a collection of Virtual IP Addresses (VIPA) to the outside world. For example, the primary site exported 2 different VIPAs, one for ....

O. Kremien, J. Kramer, and J.N. Magee. Scalable and Adaptive Load Sharing for Distributed Systems. IEEE Parallel and Distributed Technology, 1(3), 1993.

....1 . Guidance as of possible use as well as error situations more often encountered in a distributed environment are also included. 1http: www.cs.biu.ac. il dsg statistical tests.html 3 We describe the use of tools for implementing the advocated analysis method on a load sharing algorithm [10] designed to be scalable and adaptive. The same tool and analysis method can serve to analyze the results of other distributed algorithms: this particular example is chosen for illustration only. The statistical analysis method advocated was already used in several cases where users could directly ....

....of full cooperation (ideal) is represented by the M M n model. This model implicitly (and unrealistically) assumes accurate system state information at no cost. Few cases were run and compared successfully to M M n results. To further illustrate our method, we select an example from the literature [10]. Results have been reproduced by both CNCSIM and NEST. This is discussed next. 4.6. Production of Initial Results Reproduction of earlier results was important for gaining more confidence in the package selected to run the algorithm under study. The results published for FLS in [10] were produced ....

[Article contains additional citation context not shown here]

O.Kremien,J.Kramer,J.Magee, "Scalable and Adaptive load sharing for Distributed Systems", IEEE Parallel and Distributed Technology, 1(3), Aug. 1993, pp. 62-70.

....errors occur. The local communication times are dominated by the time taken to perform a light weight thread context switch on the SPARC architecture. Local Remote Message Size(bytes) Test Synchronous X.out(M) X.in(M) Asynchronous X.send(M) X. in(M) 1 100 1000 1 100 1000 118uS 126uS 197uS 1.89mS 2.05mS 3.01mS 121uS 131uS 182uS 0.98mS 1.16mS 2.17mS Table 1 Port communication performance In this section, we have illustrated how communication is supported in Regis by C objects. In particular, we have used the Regis port object as an example. Ports can be used to transfer ....

....occur. The local communication times are dominated by the time taken to perform a light weight thread context switch on the SPARC architecture. Local Remote Message Size(bytes) Test Synchronous X.out(M) X.in(M) Asynchronous X.send(M) X.in(M) 1 100 1000 1 100 1000 118uS 126uS 197uS 1. 89mS 2.05mS 3.01mS 121uS 131uS 182uS 0.98mS 1.16mS 2.17mS Table 1 Port communication performance In this section, we have illustrated how communication is supported in Regis by C objects. In particular, we have used the Regis port object as an example. Ports can be used to transfer messages with ....

[Article contains additional citation context not shown here]

O. Kremien, J. Kramer and J. Magee, "Scalable and Adaptive Load Sharing for Distributed Systems", IEEE Parallel and Distributed Technology , 1 (3), August 1993, pp. 62-70.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC