Romex Software Forum

I have a 2 Gb L1 (R/W) and 200 Gb L2 (R/W). In defer write, I have Flush L1 to L2 deselected. I notice that with this setup between 15% and 25% of Total Write are written to L2. Why does it write to L2 in this setup? Is it because It found those blocks in L2 and quicker to update direct to L2?

Even with a 32 Gb system and after a while of general use so that Windows cache gets populated over time, PrimoCache still does a lot of reading. Is the Windows cache that inefficient? Is it possible in a future build to also capture the value for Windows cache I/O activity, so users can know just how much as a percentage of all system I/O that PrimoCache is working? It would be an interesting metric for the users to know.

RobF99 wrote: ↑Thu Dec 23, 2021 12:17 pm I have a 2 Gb L1 (R/W) and 200 Gb L2 (R/W). In defer write, I have Flush L1 to L2 deselected. I notice that with this setup between 15% and 25% of Total Write are written to L2. Why does it write to L2 in this setup? Is it because It found those blocks in L2 and quicker to update direct to L2?

Two reasons: 1) When L1 is full of deferred write-data and triggers "urgent" writing-back, L2 will accept new write requests. 2) When data blocks are already in L2, PrimoCache will do a direct update to L2, as you have thought.
If you don't want L2 for caching writes, just set L2 100% read.

RobF99 wrote: ↑Thu Dec 23, 2021 12:17 pm Is the Windows cache that inefficient?

Windows use free RAM to cache the whole computer. With limited cache space, a large file read/write will easily evict old cached data.

RobF99 wrote: ↑Thu Dec 23, 2021 12:17 pm Is it possible in a future build to also capture the value for Windows cache I/O activity

Possible, but at a very low priority.