How does KVM cache work?

[dchiesa1]

Did you read the documentation? What part did you not understand?

Maybe you could provide more background on specifically what you'd like to clarify.

[sillan_dt]

You have below simple things to understand.

1. You specify the cache expiry in the policy, that holds the KVM cache till that long, or put operation is done.

2. If a put operation happens before the cache expiry also, that will refresh the cache.

3. If no put operation is done and the cache expiry time is completed, then the KVM from cache will be gone.

4. When again you will do a get operation on that KVM, the particular total KVM, including other entries in it will be cached.

[Denis_KALITVI]

I think its a little deeper than that - if you have number of RMP's then cache synced to first RMP that hit this KVM etc. - not sure of details , but I think cache (no KVM) work like this,(OPDK, right)

[sillan_dt]

If you will go little deeper, then KVM cache is stored in L1 and L2 both. When it is in L1 it is specific to a message processor. If the same message processor takes the next call, it will take from L1 cache. If first time one message processor calls then obviously L1 cache wouldn't be there. But instead of going to the DB it will go to L2 cache and take from there.