Configuration: Windows Server 2012 R2, vomume tested;RAID 5 (4+1), LSI MegaRaid 84016E, SATA II 256 MB W/BBU. The benchmark tests looked very promising. I created a Cache task for the volume using Read & Write. I used 4096 of the OS memory for level 1. There is no level 2. I tried all the block sized and increased the defer write time, 10, 300 and 1800.
Scenario: 500GB of large files copied from NAS. The result is consistently that the copy without the cache enabled is faster. I would expect that with a SATA II/RAID 5 write penalty configuration that more system memory would speed up the transfer. I guess I am missing some principle of the Primocache. Can anyone help?
OS write performance is better L1 Cache
-
BabyfaceMcGill
- Level 1
- Posts: 1
- Joined: Thu Jan 01, 2015 7:38 pm
-
InquiringMind
- Level SS
- Posts: 477
- Joined: Wed Oct 06, 2010 11:10 pm
Re: OS write performance is better L1 Cache
Welcome to the forums BabyfaceMcGill,
Caching only provides a performance benefit when re-reading (or re-writing) data, not on first read/write. The reason that many benchmarks report a major improvement is because they repeat reads/writes using the same data. PrimoCache should provide an improvement in day-to-day use, but far less than most benchmarks would suggest. Additionally for NAS drives, network issues (bandwidth limitations, driver conflicts) may come into play.
However another possibility is lack of system memory if you have allocated too much to PrimoCache - how much RAM does your system have? This would seem the likeliest explanation for poorer performance with PrimoCache enabled (for an "ideal" L1 cache size, find your maximum memory usage, subtract that from total memory, remove 1-2GB for a safety margin and allocate the remainder to PrimoCache).
Caching only provides a performance benefit when re-reading (or re-writing) data, not on first read/write. The reason that many benchmarks report a major improvement is because they repeat reads/writes using the same data. PrimoCache should provide an improvement in day-to-day use, but far less than most benchmarks would suggest. Additionally for NAS drives, network issues (bandwidth limitations, driver conflicts) may come into play.
However another possibility is lack of system memory if you have allocated too much to PrimoCache - how much RAM does your system have? This would seem the likeliest explanation for poorer performance with PrimoCache enabled (for an "ideal" L1 cache size, find your maximum memory usage, subtract that from total memory, remove 1-2GB for a safety margin and allocate the remainder to PrimoCache).