Perf and Sizing Data from MSW Enterprise Search
Great Perf and Sizing Data from Lauri from the MSW Search Team. I'll send a link when their white paper goes live. This data came from a recent thread between Lauri and a customer asking some sizing questions. Understand this is raw and off the cuff. Just like you like it I hope.
Questions to be clarified with MSFT in verifying suitability of approach
1a. Estimate of Indexed Content
12+ TB in SharePoint Content Databases (mix of 2003 / 2007), unknown size outside of this environment
b. Total size of the index
• SSP search database ~200GB
• SSP profiles database ~30GB
• Index size on disk ~75GB
c. Total number of objects
15.4 million - will likely top 20M once we get the full crawls run on the new RTM SSP. The http://team farm alone is up to 9.4 M items for example.
e. Typical 'real world' query response time from this implementation
~2 seconds, although the product group is looking into ways we can optimize this for our environment.
f. Description of the hardware configuration including number of machines, RAM and Processors
For the Redmond search deployment:
2 Web Front-End / Query Servers
• Single dual-core processor
• 4GB Memory (adding 4 more per machine soon)
• Windows Server 2003 (64-bit)
• Windows NLBS
OS – C:\ 50GB
Program files D:\ 18GB
Utility: E:\ 67GB
Index: I:\ 300GB
1 Indexing Server
• Four dual-core processors
• 16GB Memory
• Windows Server 2003 (64-bit)
• OS – C:\ 50GB
• Program files D:\ 18GB
• Utility: E:\ 67GB
• Index: I:\ 300GB
• Index Dump (where we put backups): 600GB
Database Back-End
• SQL 2005 Cluster
• Windows Server 2003 (64-bit)
• Four dual-core processors
• 8GB Memory
• Different drives arrays for the search Db, for the SSP Db, for each content Db, for the logs and for the SQL out of box dbs (temp, master, etc).
g. Description of the type of content permissioning used for this environment
Authenticated Traffic only. Windows... NTLM or Kerberos.
2. Please indicate all known scaling metrics such as:
You’ll want to get this data from the capacity planning doc.
We’re ready to scale to 25+ million on our current configuration. We may be able to hit 50 million on this hardware as well, but are not there yet.
