0

I've got the next setup:

  • Proxmox 7.2
  • CEPH 16.2.9
  • K3S v1.23.15+k3s1
  • CEPH CSI v3.7.2

CEPH using as RBD-storage for QEMU images and K8S PVC. When I do disk benchmark in QEMU I've got the next results:

Name Read(MB/s) Write(MB/s)
SEQ1M Q8 T1 16122.25 5478.27
SEQ1M Q1 T1 3180.51 2082.51
RND4K Q32T16 633.94 615.96
. IOPS 154771.09 150380.37
. latency us 3305.38 3401.61
RND4K Q1 T1 103.38 98.75
. IOPS 25238.15 24109.38
. latency us 39.06 40.30

But when I do the same in K8S results worse

Name Read(MB/s) Write(MB/s)
SEQ1M Q8 T1 810.36 861.11
SEQ1M Q1 T1 600.29 310.13
RND4K Q32T16 230.73 177.05
. IOPS 56331.27 43224.29
. latency us 9077.98 11831.65
RND4K Q1 T1 19.94 5.90
. IOPS 4868.23 1440.42
. latency us 204.76 692.60

I'm using writeback cache for QEMU. If i disable cache the results looks like K8S. Is there similar writeback mechanism in K8S or CEPH CSI?

chs_voks
  • 33
  • 1
  • 5

1 Answers1

0

kube-fledged is a kubernetes operator for creating and managing a cache of container images directly on the worker nodes of a kubernetes cluster. It allows a user to define a list of images and onto which worker nodes those images should be cached (i.e. pulled). As a result, application pods start almost instantly, since the images need not be pulled from the registry.

kube-fledged provides CRUD APIs to manage the lifecycle of the image cache, and supports several configurable parameters to customize the functioning as per one's needs. Go through the source documentation for more details. (Source: kube-fledged github pages)

Kranthiveer Dontineni
  • 167
  • 8
  • Unfortunately, this solution is for image caching not for containers i/o – chs_voks Jan 09 '23 at 19:08
  • Then try to use local ephemeral storage for caching your container images, for more information refer to this official [documentation](https://kubernetes.io/docs/concepts/configuration/manage-resources-containers/) – Kranthiveer Dontineni Jan 10 '23 at 03:17