| 摘要 |
PURPOSE: A method for analyzing a server bottleneck in real time is provided to search a bottleneck portion with accuracy through an exact performance analysis according to resources, an analysis method from a low speed resource to a high speed resource sequentially, and method for analyzing an influence between each resource and to minimize a load of a server by accessing a memory image of a kernel directly or using a spin lock using a system call, a shared memory, and a mutual exclusion. CONSTITUTION: A server information collection and performance analysis agent module is installed in a server. A critical value with respect to a usage of each resource is set according to use objects of the server. Threads for collecting state information of each sever resource are created. CPU information, a virtual memory information, disk output information, network input/output information, process information, user information, file system information, and hardware information are collected in real time from the created thread. The current state value with respect to the collected server resource is compared with the critical value. A performance analysis is executed in order of network input/output information, disk input/output information, virtual memory information, and a CPU information, that is, from a slow band width to a rapid band width. A performance analysis result and state information of each server resource are processed and output as a graph form and a chart form.
|
