As a big data application software,distributed database needs limited system resources to complete complex concurrent computing and storage tasks.Massive data computing also requires a lot of memory,and in order to isolate multi-task resources,it is necessary to limit memory for different task groups.When out of memory(OOM),the operating system defaults to forcing to kill processes as a high-risk behavior.Therefore,this research starts with a simplified application,explores a method of process freezing and unfreezing based on the Linux kernel control group(cgroup)feature when OOM,and then applies it to the open-source distributed database Greenplum Database.
维普
万方数据