VMware deduplicated Datastore with StarWind

[link855]

Good morning, I’m trying in a little lab StarWind Deduplication LUN feature with vSphere 6.
On a single Windows Server 2012 R2, I have install StarWind and create a 400GB deduplicated iSCSI device and connected to a ESXi 6.
I have moved several VMs on this StarWind Datastore and after a comparison between VMware Datastore space usage and StarWind files, I see:
VMware Datastore 380 GB used, StarWind summation size 180 GB with deduplication ratio 2.25.
I’m very satisfied for these tests with this product, simple and efficient!
I have a couple of questions to deepen the capabilities of this product to successfully address any customers requests:
- I have created a deduplicated device of 400 GB and with deplication the space physically used is only 180GB, but VMware datastore seems full; how I can continue to use from VMware this datastore until it will be really full (physical size 400GB)?
- I run comparison performance test between StarWind deduplicated device and thick device, both with 4 GB L1 cache write-back: the deduplicated device IO speed results half compared to seped of thick device. How can I improve performance of thin deduplicated device?

Thank you in advance

[Tarass (Staff)]

Hi Link,

The problem of VMWare working with thin provisioned devices is weird but known
For now, the only reasonable way would be to create a larger LSFS device initially or create a second one near existing. In early 2016 we are going to release an LSFS-expand feature so you will be able to expand your existing LSFS devices.

[link855]

Hi Link,

The problem of VMWare working with thin provisioned devices is weird but known
For now, the only reasonable way would be to create a larger LSFS device initially or create a second one near existing. In early 2016 we are going to release an LSFS-expand feature so you will be able to expand your existing LSFS devices.

Thank you Tarass for your reply; what about performance comparison between Thick provisioned and LSFS volumes?

Greetings

[Tarass (Staff)]

Testing performance depends on testing pattern/scenario. Try 4k random blocks (if you are virtualized, it's the worst scenario) and see what happens

[fbifido]

I have moved several VMs on this StarWind Datastore and after a comparison between VMware Datastore space usage and StarWind files, I see:
VMware Datastore 380 GB used, StarWind summation size 180 GB with deduplication ratio 2.25.

Which version of Virtual SAN, are you using ?
What's the actual lsfs disks space used ?

if it's 180 GB, then that means the 3.5x storage problem was fixed.
https://forums.starwindsoftware.com/vie ... =15#p23200

Thanks.

[Tarass (Staff)]

Thank you all for valuable contribution. Anything else I can help here with?