Dedupe question

Fri Nov 04, 2016 12:15 pm

Hello xy67,
I found that there were issues with this switch before: https://supportforums.cisco.com/discuss ... loss-speed, but, unfortunately, does not have any idea which setting could fix the issue.
As for LSFS question, with deduplication enabled, with 3 x 50 GB VMs, the 150 GB datastore will be full, but space, occupied by LSFS device, should be about 50-60 GB. At the same time, it could occupy more space because of data type and deduplication ratio.

The MPIO role should be installed in Windows in case if you are connecting the targets in Windows itself. For compute and storage separated scenario, there is no need to enable MPIO.

Usually, StarWind devices show in ESXi as Non-SSD. You can check it from GUI or alternatively you can run "esxcli storage core device list" command to list all disk devices and their properties. There should be the line "Is SSD:" which reports about disk type.

xy67 · Fri Nov 04, 2016 12:55 pm

Thanks Michael

It is a problem with this switch. Something to do with port/packet buffer size. I have since connected my ESXi host to the SAN directly via "crossover" cables (bypassing the switch) and it is much better so far.

I must be dumb because I'm still not getting the sizing of the LSFS device with dedupe

If the 3 x 50GB VMs fill up the 150GB datastore then ESXi won't let me create new VMs because it thinks the datastore is full surely? Even though there is free space from Starwinds point of view (with dedupe) ESXi still thinks the datastore is full doesn't it?

Fri Nov 04, 2016 6:54 pm

Yes, you are correct , the ESXi won't let you create new VMs because it thinks the datastore is full. From another point of view, you have saved the space on D:\ drive, where LSFS device located. It gives you a possibility to create other devices there or use it according to your needs.

xy67 · Mon Nov 07, 2016 1:51 pm

What is the recommended write back cache amount to set when creating an LSFS device with dedupe? I've been using 4GB but is this too much or little?

Tue Nov 08, 2016 5:39 pm

Hello xy67,
The base recommendation is to put 1 GB of L1 cache in Write-Back mode per 1 TB storage capacity. Basically, configuring bigger cache size will give you better performance.

xy67 · Tue Nov 08, 2016 7:40 pm

Is it possible to change the write back cache amount from 4GH to 2GB without recreating the device?

Also does Starwind Virtual SAN support IPv6?

Thu Nov 10, 2016 10:27 am

Hello xy67
Yes it is possible to change cache amount without recreating the device.

https://knowledgebase.starwindsoftware. ... -l1-cache/

IPv6 is not supported by StarWind

Have great day.

xy67 wrote:Is it possible to change the write back cache amount from 4GH to 2GB without recreating the device?

Also does Starwind Virtual SAN support IPv6?