vSAN sizing calculator: what about free space

[Luca82]

Hi everyone. I'm doing a sizing for a customer, using the sizing/tco calculator tool

Broadly speaking, he needs a 2 nodes (not ROBO, but 2 nodes plus witness virtual appliance) cluster, with more or less 30vm and a total of 1,7 TB of vmdk.

Now, suppose to use FTT=1, I think that 1,7*2= 3,5 TB....plus 10% for cache and other files.....I suppose that 4TB should be a good results (2TB per node).

I've inserted all the data inside calculator, but I see that the calculator oversize the environment and suggest a lot of "additional free space" but I don't understand where I have said that I could need additional free space!

Could you help me to understand?

[img]http://imagehosting.biz/images/2017/04/05/2017-04-0511_25_06-vSANTCOandSizingCalculator.png[/img]

[Luca82]

[admin]

Greetings!

Could you please share the screenshot of "Sizing Inputs" page as well?

Here is how I entered info on "Sizing Inputs" page based on your query:

Sizing Results:

Cheers!

-Shivam

[Luca82]

Thank you very much for your answer.

This is my sizing

[TheBobkin]

Hello Luca,

Okay, my initial attempts also showed similar incorrect advised configurations.

From attempting it myself a few times and looking at your output it is suggesting/forcing multiple inputs which are causing the end-result being vastly over-sized and inaccurate:

- You are probably unlikely to get 4.00x Dedupe ratio (though this depends on the composition of the VMs, if they are full clones as per your input and they are mainly OS not user data then sure you may get somewhere near this).

- On the 'Ready-node/build your own' page it is applying use case profile of 12TB per node

- As this will be a 2 data-node cluster (Direct-connect set-up, correct?) you need to get this guide to allow for this and stop forcing 3 data-node as minimum ('Nodes Required' shows 3 at bottom of page on your images), I managed to get this showing 2 nodes by specifying using ROBO (I know you are not using ROBO but I don't think this thing knows about Direct-connect 2-node clusters as it was made before these existed) and in reality the data placement is the same if all FTT=1 Objects and 2 data-nodes.

You can see from your results page it is calculating for a 3 data-node cluster each with 12x800GB capacity drives totalling 28.8TB raw capacity.

Are these View VDIs? (as I saw you specified 'Full clones')

If so this might be helpful (a bit antiquated but gives some good sizing assumptions assuming concepts remain the same nowadays):

http://www.vmware.com/content/dam/digitalmarketing/vmware/en/pdf/whitepaper/products/vsan/vmw-tmd-vi...

If getting the auto-sizing guide to play ball is not feasible (I will have another attempt myself) I would probably advise creating your own presentation material after reading the relevant sections of this:

http://www.vmware.com/content/dam/digitalmarketing/vmware/en/pdf/products/vsan/virtual-san-6.2-desig...

There are some basic Excel based-sizing resources online such as this that might be of some use (but not so much with 2 data-node clusters I think):

http://vmusketeers.com/vsan-6-2-resources-calculator

Bob

-o- If you found this comment useful please click the 'Helpful' button and/or select as 'Answer' if you consider it so, please ask follow-up questions if you have any -o-

[TheBobkin · Accepted solution]

Hello Luca,

Okay, I figured out the main problem with why it is over-sizing total output:

- If you do not select a number of 'Flash Cache devices in Server" (cache-tier SSDs) or 'Persistent Disks in Server' (capacity-tier drives) it automatically uses the Ready Node configurations as per the column on the far left (12x capacity disks in your screenshot).

I selected the same settings as all of yours but with the addition of selecting ROBO (for the reason in my last comment, as this is the only way to get it to do 2 data-node cluster) for each node I selected 2x 960GB capacity-tier SSDs, 1x 400GB cache-tier SSD.

I configured using these correct and got the following (again, not sure if you will get 4.00x Dedupe ratio but that is what I input here as you did before):

Bob

-o- If you found this comment useful please click the 'Helpful' button and/or select as 'Answer' if you consider it so, please ask follow-up questions if you have any -o-

[TheBobkin]

Additionally you could even just use 200GB cache-tier SSDs instead of 400GB as used in the last shown set-up, these would meet the advised 10% cache:capacity ratio.

But if you think you might add additional capacity-drives at some later date you should consider using 400GB cache-tier drives to allow for this expansion to be easier.

Whether this might be needed or not may also depend on the dedupe ratio achieved and this of course depends on the data-set of the disks and their commonality with each other.

Another option instead of using the 2x960GB capacity-drives per disk-group as in my example, you could use 3x600GB  capacity-drives (*slightly* less space) which may be more cost-effective and allow for further striping of Objects.

Bob

-o- If you found this comment useful please click the 'Helpful' button and/or select as 'Answer' if you consider it so, please ask follow-up questions if you have any -o-

[Luca82]

Thank you very much for your detailed answer