0

I'm working on a VM which regularly hits 100% CPU load. It's hard to be sure that the running processes itself are particularly heavy, but I have a suspicion that the CPU's may be outdated, which is what I'd like to get some feedback on here.

The machine runs Jenkins and a small webserver. There's also Sophos and Tanium anti-virus and endpoint security running. All-in-all nothing to extraordinary, I think.

The VM is powered by two Xeon E5-2698 v3 @ 2.30GHz. There's also 16GB of RAM of which about half is used.

Could a dual E5-2698's be considered a low-spec VM in 2022?

svenema
  • 101
  • 3
  • 2
    There's no *objective* answer as to how fast a CPU is needed. Does the VM work for what you use it for? If so, the CPU is adequate. – vidarlo Jun 03 '22 at 07:55
  • Check the filesystem for any problems. A problematic filesystem, aka a harddisk drive which is going to die soon, is often using massive amounts of CPU (kernel) time. – paladin Jun 03 '22 at 09:29
  • 1
    Hitting 100% CPU utilization on a VM simply means that the load required is equal to or higher than the CPU resources available to that VM. By itself it says nothing about the actual instructions per second capability required to run the load at a pace that’s an acceptable trade off between price and performance for your business. – Mikael H Jun 03 '22 at 09:29

2 Answers2

0

According to Passmark CPU benchmarks a Dual socket system with 2 of the E5-2693 v3 CPUs is slower than a single AMD Ryzen 5 5900 in this comparison.

While those synthetic benchmarks are only to be used as a bit of a guide.... Given the price tag of the faster AMD Ryzen 5 5900X, is about $370 USD at the moment, and most of all, because the E5-2693 v3 was released in 2014, it's likely time to upgrade before the old hardware just dies of old age.

BeowulfNode42
  • 2,615
  • 2
  • 19
  • 32
0

Really depends. I'd really like to see what the VM is doing and how it's configured.

Check if you're running into bottlenecks. I'll spray a few ideas to check (also they'll contain Linux-centric stuff, so some things may not apply):

  • to few vCPU cores – you're just vCPU bottlenecked, either too much processes (high load average or excess context switches), or one process hugs too much single-core performance (would require more debugging),
    • which can lead to too much webserwer connections, Jenkins tasks, but this is crystal ball-level for now,
  • high iowait – I/O bottleneck (disks, network, etc.),
  • vCPU pinning – having too much processes
  • VM configuration – last shot, as by default this shouldn't be an issue.

Yes, those are very broad, but can help narrowing down if you have an software(-ish) issue that might be saved and avoid you a new CPU.

Dźwiedziu
  • 1
  • 1