VSANchat

No guarantee of performance improvement, but great from an availability perspective (with a side effect of "possibly" improved network performance)

@CormacJHogan We are using two 10GB NICs per host in a single vSwitch in active passive. WOrks for us, but only provides fault tolerance. THere are VMware docs out there, but a simple, concise best practices doc would be great.

@num1k got it - heard this before too.

@CormacJHogan Is dual NICs in active/passive config on a single vmk/vSwitch the best way to utilize 2 dedicated VSAN NICs/server? Can active/active provide the fault tolerance & improve total available throughput?

one doc mentioned VSAN networking has the same requirements as vMotion. We use 2 vnics per host for vMation. That;s not possible with VSAN. Failover is good, but does not require 2 VLANS.

On this note, is there a best practice for Network IO Control profiles for use with other shared NICs and traffic types?

@num1k So there's no guarantee that any network configuration will improve performance. So really, there are a couple of different ways to configure VSAN networking

@v2Mike Not really Mike. The only traffic that we recommend throttling with NIOC is vMotion which is notoriously greedy

@CormacJHogan We are finding individual spindle speeds are our performance limiter. But just turn up the stripe width, VM can push more IOPS. #itgoesto11

@num1k During cache destaging? Or are you having read cache misses?

You can do it one disk group at a time. Evac disk group, remove it, create new AF disk group - rinse and repeat

how many hosts do you have?

@jasemccarty currently 4 we are looking to add a 5th

@CormacJHogan thanks I tried this in the HOL and figured it would be the easiest

@CormacJHogan I'd think that a new Cluster would have to be setup, as VSAN is either Hybrid or All-Flash (6.0) but cannot contain both configurations.

a couple hosts could be "brought out" and setup in a 2 node briefly, as you rolled through changing the config.

@jasemccarty good point about support - I "think" it will be ok (done it myself) but best to be on the safe side

@CormacJHogan safe & supported are good.

VSAN does use thin provisioning. A question I would ask, is what is your FTT setting?

FTT is 1 (granted drives are 40GB on each node) with three nodes. NSX Manager is 60GB (but 3GB if thin).

Those OVAs are probably already thick when they were created. I don't think we can do anything about that. They'll roll out thick IIRC

I think @CormacJHogan is correct there.

to be able to use Stretched Clusters or ROBO, you would have to.

Well... I do get paid for these things... :)

Also new health check features and in 6.0u1

SRM can be used to protect the Stretched Cluster to a 3rd sight, but not between the 2 sites. Add vSphere Replication from the Stretched Cluster to the 3rd site (with VSAN on the backend) and you can get a 5 min RPO.

A blog by @DuncanYB was written on this recently of having Stretched Clusters and SRM: https://goo.gl/yn0ob...

seemless flash acceleration!

low cost (next to NAS/SAN), speed, and storage profile flexibility

Health Check enhancements are nice to see. Not having to deploy advanced monitoring products right away speeds up deployment while still having insight into storage. The better overall performance is great too.

@brugh What do you mean by this? The flash acceleration is seemless imo.

@RyanGarvelink exactly ;) that's my favorite benefit

@brugh Lol :)

Stretched Cluster for VSAN is definitely here ... you can view some great info here http://vmw.re/1OcXIZ...

yes it is! Check this out for details: http://www.vmware.co...

vSphere 6.0 Update 1 includes Virtual SAN 6.1. It was released on September 10th.

Thanks much :)

Not at present. There needs to be VSAN connectivity between three hosts (witness too), and you need a switch to do that. However, this isn't the first time we've heard this, especially for ROBO, so we're looking into ways to do it

Great, thanks for the input. Makes sense.

12000 user Lync platform

ROBO sites where they have old physical servers ready to refresh. Rather than buy new SAN for ROBO, vSAN works perfectly.

fairly pedestrian deployment, backing our new VDI environment. I assume this is one of the top use cases at this time.

maybe a dedicated Oracle cluster so you don't have to worry about Oracle licensing...

We've seen some great comments on #VSAN for ROBO and some VDI questions ... how else are people using #VSAN (or seeing it used)?

I hope a refresh for EVO RAIL with 2 node setup will follow

@VmAlchemist VDI is definitely a popular use case - the cost, predictable scaling and performance make that a great fit ... are you deploying on all-flash or hybrid?

@MichaelJHaag curious about mission critical apps on #VSAN too

@vGazza VSAN 6 is ready for mission critical apps!

replaceing 96 TB iSCSI SAN with 5 node VSAN

@LTVSAN no doubt about it. Curious about references.

@MichaelJHaag We're deploying hybrid -- 2 disk groups per host --- 1.6TB SSD and 7x2TB SAS disks

@vicenac curious how you sized it; does it look anything like what http://virtualapplia... would advise?

We're seeing lots of adoption of Virtual SAN for VDI, which many customers see as Mission Critical.

@vGazza Here is an example of T1 apps (Oracle, SQL server and a bunch of medical apps) on VSAN in healthcare: http://www.vmware.co...

@jasemccarty very true. Nothing more critical than end user computing platform.

@vmw_pooja thanks! Great use case!

Replaced aging iSCSI/FC SANs at multiple sites with VSAN clusters. Better performance, fault tolerance, cheaper, more flexible, more GB/dollar, easier to deal with.