VSAN ESA STRECHED CLUSTER ARCHITECTURE

[Ferhat_Tuna]

Hi,

I am designing a vsan esa project to be used in 3 sites.

 

The topology consists of 21 nodes in total and each site consists of 7 nodes as independent clusters.Desired net capacity for each site is  FFT2-RAID 6 312 TIB  (excluding operational reserve)

 

You can think of 2 sites as active-active streched clusters and 1 site as disaster recovery.

 

Customer wants to see 50,000 iops value in under 1ms with 32K blocks size in 67% read profile and 33% write profile.However  vSAN ESA sizer  doesnt  allow over 16K block size (attached Q1 JPEG ), I would be very happy if I get find out why.

 

Also, when I activate the streched cluster in the sizer, the operational reserve becomes closed (attached Q2 JPEG). If I open the operational reserve, the streched cluster tab becomes closed. What is the reason of this?

 

How can I calculate operational reserve in a stretched cluster architecture? If you have any suggestions, it'll be very important for me.

 

There is no streched cluster tab in Quick sizer, in addition, I wanted to consult you because when I activated the operational reserve in detailed sizer, it disabled the streched cluster tab.

 

vsan ESA is a new technology, I don’t want to encounter any surprises during the installation phase.If I get help as soon as possible, I’ll be appreciated.

 

Thanks

[Duncan Epping]

I am not sure I understand the question, seems like we are mixing something here. 

1. You state you want a stretched cluster, but then talk about 3 sites with independent clusters? 
2. Operational Reserve in combination with Stretched vSAN is not supported (https://core.vmware.com/blog/understanding-%E2%80%9Creserved-capacity%E2%80%9D-concepts-vsan)
3. 50k IOPS in under 1ms sounds like a normal cluster? vSAN probably isn't your limitation here, but potentially networking?
4. vSAN ESA Stretched requires 35Gbps between sites, is that available?

[Ferhat_Tuna]

Hi again,

Thanks for response

When I say independent,  I mean is that the system will be used in 3 different sites. I expressed it wrong.

Capacity is requested to include operational reserve and I need to document it. At this point I am having a problem.

There is no problem in the network infrastructure, but as I mentioned before, the performance values ​​that I mentioned for 32K blocks are requested. Can I have your comments on this subject?

[Duncan Epping]

So, operational reserve cannot be enabled on vSAN Stretched, I would probably take the percentage into account which matches your host configuration closest as demonstrated in this post: https://core.vmware.com/blog/understanding-%E2%80%9Creserved-capacity%E2%80%9D-concepts-vsan

Roughly 10-15% you normally should be good. 

 

[Duncan Epping]

When it comes to the performance, under 1ms is a big ask for a stretched cluster, from a storage perspective this should be doable with vSAN ESA and NVMe for sure. But the network in between locations, and in each location, is going to be the issue most likely. As each IO will need to be acknowledged from both locations it means you can only really have 0.5ms latency between location A and B, as 1ms is roundtrip I am assuming? And actually it will need to be less, as 0.5ms doesn't leave room for processing the IO. So instead of 1ms RTT for the network it probably should be around 0.5ms RTT instead. So the question probably is not if vSAN can meet those requirements, but can the network to start with?

[Ferhat_Tuna]

I definitely understand that it is related to the network. My main question is that I cannot show these performance values for the 32k block sizer output before the bidding.

And in sizer, random %33 read 67% write option appears for 16k size.

The main problem I have is why we can't show these values in 32k block size.

Your opinion on this matter is very valuable to me.

 

Thanks

[Duncan Epping]

No idea why we have 8k or 16k as an option and nothing larger, will ask the development team what the reason is behind it.