What is the best architecture for a SharePoint 2016 farm having a content of13 TB?

The idea is to get redundancy and performance using this search architecture.

Are 2 servers helping in this case with 2.5 TB disks to store index files with its replicas on the other server?

Is this option feasible?

  • 1
    What is the estimate for number of item and documents? – Matthew McDermott Jul 15 '16 at 14:26
  • At this moment, I know only about the content size but not the number of items or documents. Is there any way to identify from content size? – scks Jul 19 '16 at 6:00
  • In the SSA there is an indexed item count. That is how you determine the scale of your implementation. Not by the size of the corpus. – Matthew McDermott Jul 19 '16 at 21:25
  • Thanks Matthew for the comments! In my case its like a greenfield installation and expecting 13 TB content in the farm. Hence, I m unable to find it from indexed item count in SSA. – scks Jul 20 '16 at 12:50
  • Can you use powershell to audit the source documents and get a count? – Matthew McDermott Jul 21 '16 at 20:54

This is very much an 'it depends' question!

Work out how many items you'll have that will be crawled. For SharePoint 2016 the maximum supported number of items per index partition is 20 million (see Software boundaries and limits for SharePoint Server 2016, so divide the number of items you have by 20 million and multiply by 2 to get the absolute minimum number of servers you should deploy in the search query partition role. Divide the servers into pairs and each pair should host a replica of an index partition, with the total index requirements being spread across these pairs of servers. Obviously you will need additional servers in your farm for other roles.

See Technical diagrams for SharePoint 2013 for a set of useful diagrams for SharePoint 2013 that can help design your search architecture amongst other things.

  • Andy, thanks for your response! Can you please let know how the number of items per index partition to be calculated. As far as I know, there will around 13 TB contents. Are there any ways to calculate the items from content size? – scks Jul 19 '16 at 5:55
  • If you cannot get an accurate assessment of the number of files, you'll have to take a representative sample and calculcate the average file size and use that to determine an estimate of the number of files that you have. Obviously the mix of files will potentially be different for different areas of the business, so it's important to try and take this into account if possible. Note that the number if items per index partition is a support boundary, so if exceeding that, the system may continue to function, but you may need to add servers/reduce the number of index items if support is required – Andy Dawson Jul 19 '16 at 8:16
  • In SP2016, all the search related services and index files will be running and stored in servers designated as Search MinRole? – scks Jul 19 '16 at 9:12
  • Yes, using the search MinRole, those servers would be running the search related services and store the index files. – Andy Dawson Jul 19 '16 at 11:59

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.