2

We have some servers CentOS 7 with 1Tb RAM for large math calculation.

When server is just rebooted, at application start memory allocation is realy fast enter image description here

When some calculations were already performed on the server (large uptime) (more then 3 hour required for memory allocation)

enter image description here

Strace log fullfilled with brk and sched_yield

brk
brk
brk
brk
....
sched_yield
sched_yield
brk
....

Sysctl: vm.swappiness = 1 Monitoring show that swap not used during memory allocation.

Is it possible to speed up memory allocation for application with some system tuning?

Improve this question
7
  • Does the allocation use huge pages? Commented Dec 3, 2024 at 3:44
  • 1
    The easiest ways to speed up memory allocation will all involve changing the way the allocation is performed; 3 hours for a single allocation is far too much. Commented Dec 3, 2024 at 3:51
  • >Does the allocation use huge pages? I dont know. How to understand this? >changing the way the allocation is performed; How to change this? Commented Dec 3, 2024 at 6:01
  • 1
    Well, the first step is looking at the code for the application; how specifically does it allocate its half-gigabyte of memory? To find out about huge pages, you can read up online; I suspect though that it either is, and your memory is very fragmented, or isnt, and your algorithm is bad. Commented Dec 3, 2024 at 6:20
  • This is vendor's application, we have no access to source code and no access to support. I have only strace output at the moment when application allocating memory. Commented Dec 3, 2024 at 8:21

0

Reset to default

You must log in to answer this question.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.