Strange results from 'evaperf vdts'

[Nelson Jeppesen]

So I've been having lots of performance issues with the EVA but HP support offers no answers.

So I'm looking at 'evaperf vdts' (Virtual Disk Transfer Size) and I see lots of 0 < 2k transferers. I'm a little confused; We see the NTFS cluster size to 64k. How can the transfer size be 2k (or less)?


ID 0 2K 4K 8K 16K 32K 64K 128K 256K 512K Ctlr LUN Node
< < < < < < < < < <=
2K 4K 8K 16K 32K 64K 128K 256K 512K INF
-- ----- ----- ----- ----- ----- ----- ----- ----- ----- ----- ---- --------------------------------------------------------- --------
59 996 0 6 1 0 0 0 0 0 0 300A SQLogs NL-SAN
59 1000 2 4 0 0 0 0 0 0 0 400S SQLLogs NL-SAN

[Nelson Jeppesen]

Re-formated to easier reading

2K 4K 8K 16K 32K 64K 128K 256K 512K INF
996 0 6 1 0 0 0 0 0 0 SQLLogs
1000 2 4 0 0 0 0 0 0 0 SQLLogs

[Uwe Zessin]

The "cluster size" or "block size" of a file system just defines the smallest allocation unit covered by a single free/used bit - it does not imply that the OS I/O size is fixed to this value and the driver is then doing blocking/de-blocking, because the application does smaller I/Os.

[Nelson Jeppesen]

So lots of 2k IO to SQL logs normal?