I haven't figured out how I'm going to best take advantage of this technology. Has anybody experimented with block sizes to find a sweet spot for mixed workloads? Most of our VM's are windows probably with the majority having drives formatted default 4k. We do have quite a few SQL vm's with 8 to 64k block sizes. Our backend storage is currently VNX with Fast Cache. The SSD's we have in our new esx boxes are SAS attached which I know are not the fastest out there but I'm hoping we can see some benefit properly done. Thanks for any thoughts.
Hi,
Something to keep in mind with vFRC is that the block size is for the application IO size, not the underlying NTFS/EXT file system formatted to a certain version.
But yes, typically MSSQL workloads use 64k IO size.
Check out my blog on sizing your vFRC cache correctly:
http://nickmarshall.com.au/blog/2013/10/31/rightsizing-vflash-read-cache-vfrc
Cheers,
Nick
Thanks Nick, that sounds more like it, don't know why I was thinking it was the disk block size.I'll check out your blog.
Nice write up. One thing I found when I started working with SSD's and ESXi was an issue with the hypervisor not being able to see the disk resource as local so you saw nothing in vcenter. I found this little tidbit that might be useful in adding to your blog
Device Display Name: HP Serial Attached SCSI Disk (naa.600508b1001c3b7559b32be63f27a8ba)
Storage Array Type: VMW_SATP_LOCAL
Display Name: HP Serial Attached SCSI Disk (naa.600508b1001c3b7559b32be63f27a8ba)
Has Settable Display Name: true
Size: 95362
Device Type: Direct-Access
Multipath Plugin: NMP
Devfs Path: /vmfs/devices/disks/naa.600508b1001c3b7559b32be63f27a8ba
Vendor: HP
Model: LOGICAL VOLUME
Revision: 3.42
SCSI Level: 5
Is Pseudo: false
Status: on
Is RDM Capable: false
Is Local: true
Is Removable: false
Is SSD: true
