In this paper, the scaling mechanism of the distributed Web
system is described.
It is confirmed that the scaling mechanism is possible
to automatically adjusting the number of working servers
according to load. Moreover, we examine response time and
running cost to confirm performance of the scaling mechanism.
As a results, both sufficient response time and low running cost
are achived.
Future works include following things.
• Evaluation with other scenarios such as changing accesses
drastically, using access logs of working Web
servers, and so on.
• Implementation and evaluation of VC-Server management
function to boot and shut down VC-Servers.
In this research, the number of virtual servers is more than
the number of CPU cores of host server but less than the
number of threads. Therefore, the cases may occur that one
server uses one CPU core and two servers share one CPU
In this paper, the scaling mechanism of the distributed Websystem is described.It is confirmed that the scaling mechanism is possibleto automatically adjusting the number of working serversaccording to load. Moreover, we examine response time andrunning cost to confirm performance of the scaling mechanism.As a results, both sufficient response time and low running costare achived.Future works include following things.• Evaluation with other scenarios such as changing accessesdrastically, using access logs of working Webservers, and so on.• Implementation and evaluation of VC-Server managementfunction to boot and shut down VC-Servers.In this research, the number of virtual servers is more thanthe number of CPU cores of host server but less than thenumber of threads. Therefore, the cases may occur that oneserver uses one CPU core and two servers share one CPU
การแปล กรุณารอสักครู่..
