-
Feature
-
Resolution: Unresolved
-
Critical
-
None
-
None
Requirement description:
As an VM Admin, I want to improve overall density. In our traditional VM environments, we find that we are memory bound much more than CPU. Even with properly sized VMs, we see a lot of memory just sitting around allocated to the VM, but not actually used. Moreover, we always see people requesting VMs that are sized way too big for their workloads. It is better customer service allow it to some degree and then recover the memory at the hypervisor level.
MVP:
- Move SWAP to beta (OCP TP)
- Dashboard for monitoring
- Make sure the scheduler sees the real memory available, rather than that allocated to the VMs.
Documents:
- Memory Overcommit Strategy (What tools we want to use): https://docs.google.com/document/d/1C6_idKxFgxOhcUBqjQg8pMY7xi3W7_LrqQL6Le3eBPA/edit#
- Higher Density Goals (What goals we have user/admin facing): https://docs.google.com/document/d/1DaaCmzR0OpnehFBVh6hiHTfQhfBLW7JZVXkv96EEkHg/edit#
- SWAP Goals (CNV's Goals wrt enabling SWAP in Kube): https://docs.google.com/document/d/1jwcx8YQ0mwotiT5GgVP4im13veYD7gPULTquxoX0aCc/edit#
- depends on
-
CNV-25724 R&D: Memory-Overcommit (validate swap)
- Closed
- is blocked by
-
OCPNODE-792 Graduate swap to Beta2 in Kubernetes
- Dev Complete
-
OCPSTRAT-138 Swap support in openshift for CNV only ( GA swap upstream to support CNV efforts)
- Backlog
- is depended on by
-
CNV-18918 Memory overcommit in Openshift Virtualization
- In Progress
- is related to
-
CNV-25725 Compressed VM Memory
- Closed