4

I have a Windows Server S2D cluster with five nodes. I have configured a 51GB CSV cache. However, after several days the server won't allocate more than about 1.5GB. I briefly saw 1.6GB. Why is the server not utilizing the available memory?

My guess is that entry time to live is quite short and instead of the cache extending the time to live to allow for the maximum utilization of the available cache size, it sticks to a fixed time span. I can think of many reasons why, such as entries that are past a certain age probably won't result in a cache hit and therefore shouldn't be cached or the resources required to manage a large cache aren't worth any supposed performance benefits. I'm really looking for a more informed explanation.

I might add that the cache hit rate is running between 5 and 55% from VM traffic.

W Lucking
  • 81
  • 3

1 Answers1

3

CSV Cache accelerates reads only, instead of LRU or MRU algorithms it has sort of an ARC-like block retention policy (but extremely badly implemented one). Making long story short - if you don’t read the same set of the logical blocks again and again to trash some threshold values - nothing will stay within CSV cache.

BaronSamedi1958
  • 13,676
  • 1
  • 21
  • 53