1 Rookie
•
21 Posts
0
1237
December 13th, 2021 07:00
High Read/Write Cache Miss Hit on All Flash Array
We have some Unity XT 880F that the storage arrays are serving to some Oracle databases.
Sometimes, we saw high Read/Write Cache Miss on the storage arrays and very large I/O block size. This situation is cause of high service time, high response time and latency of LUN. Why the arrays are applying I/O on disks directly and didn't use SP Cache?
There is no issue on storage array but latency is too high on All Flash Storage!
No Events found!
Davood Teymoori
1 Rookie
•
21 Posts
0
May 13th, 2022 05:00
The issue was related to “max_sectors_kb” value on Linux.
I think that EMC should add recommendation value to best practices or connectivity guide for EMC Unity.
Also read my blog post about that: I/O Block Size Best Practice on EMC All Flash/Hybrid Flash Array and Linux (Physical And Virtual) - Davoud Teimouri - Virtualization and Data Center
cadencep45
3 Apprentice
•
318 Posts
0
December 23rd, 2021 04:00
when you say sometimes..
Have you a test load to regenerate on demand issue ?
Have you opened a case with EMC ?
Having a reproducible test case will assist in determining is it the oracle database in isolation or due to a mix of workloads.
You mentioned cache misses are an issue. Providing a reproducible test case allows you to isolate database IO from other sources to determine is database traffic in isolation sufficient to create cache misses, which is normally caused by random IO rather than full table reads.