Sizing a large sharepoint environment

[ Edited ]


I'm currently creating a design for a very large sharepoint environment.

There seems to be very little documentation for large designs.

I know already this :

  • system db's - data + logs : separate lun | separate volume (lun 1 | vol 1)
  • temp db's : separate luns for each tempdb | luns can share same volume (lun 2,3 | vol 2)
  • Non-content db's - data : separate lun | separate volume (lun 4 | vol 3)
  • Non-content db's - logs : separate lun | separate volume (lun 5 | vol 4)

But for the content db's, it becomes more complicated :

The limit is 200GB, including blobs (regardless if they are remote or local)

Assume were are using Remote Blobs (RBS)

Assume the metadata/blob ratio = 20% / 80%

=> content db's can only be 40GB (with 160GB blobs)

Assume I have to size a 20TB sharepoint , I would have 4TB databases and 16TB blobs.

The 4 TB databases would be split up in 100 x 40GB content databases. (lun 6 - lun 105 | vol 5 - vol 104)

Asume 25% logs => 100 x 10GB logs (lun 106 - 205 | vol 105 - vol 204)

If I put each 40GB content database in its own lun within its own volume and I do the same for the logs, I would create 200 volumes.

If I was to have this environment in acceptance & in production, this is 400 volumes, coming very close to the 500 volume limit.

What would be the recommendation ?  place multiple luns in 1 volume ?  Or place multiple db's in 1 lun ? 

What about the 16TB cifs share for the blobs ?  Will this become a large file count issue ?

Thanks for any help.


Re: Sizing a large sharepoint environment


I will answer my own question, for future readers :

Don't put multiple large databases in one lun => restore will be a streamed copy restore

It's no problem to put multiple lun in one volume => as long as they have the same snapshot and snapvault schedule.

Fast lun restore is still possible using the lun clone split method.

So for my design : I'll put all content database luns from 1 farm in one volume, as they have the same backup schedule.

I'll do the same for the log files.