Notes from Tech Ed 2011 India - 1

"Design,Performance and capacity factors for successful intranet and internet SharePoint sites" session was presented by Sanjay Narang from MCS and following notes were taken from the session. All credit goes to Sanjay for enlightening the full house.

RPS- Request per second 
An indicator to express the number of request served by the web farm irrespective of size of the request

Points to note while tracking the load of the server

1. All Users - Total user base of the application
2. Active users - Total users who are currently using the application
3. Concurrent users - Total users who are currently using the same functionality concurrently(Given the nature of HTTP's disconnected architecture this'll be very few)

(Image from Microsoft Technet)

RPS Calculation

1. Take IIS logs from all the servers in web farm
2. Take from a peak hour traffic of a day
3. Use tools like Log parser to query the log files collected.(how to use Log parser ?)
4. Avoid requests for for image,CSS,JS while calculating RPS.
5. Avoid Http 401's also

Decision point

1. Average RPS on a day
2. Average RPS in  a peak hour
3. Max RPS in a peak hour
4. Average daily concurrent users
5. Peak concurrent users in peak time

Dataset

The size of data stored in a system is called as Dataset, typically in SharePoint world we call it as Content Database and its size

Points consider while planning for Dataset

1. Content Database size
2. Temp DB Size
3. Transaction Log size
4. Space required for full backup
5. Number of documents and versions in content db's
6. Number of metadata associated to the documents
7. Number of other list items (need to calculate the size of the list item too)

Performance and Reliability

1. Server availability - Overall uptime of the system
2. Server responsiveness - A Farm's time taken to serve the request
3. System resource utilization - CPU utilization and available memory

Boundaries and limits

When you plan for the capacity planning and management, it is essential to keep a tab on the product boundaries and limits

1. Content Database size - 200 GB
2. Site collection size  - Max 100 GB
3. Site Collection per Content database - 5000 Max/2000 recommended

General recommendation of Servers

• One WFE server per 10k users
• One SQL instance per 4 WFE servers
• 3-5 WFE Cores per Sql Core
• One DC per three WFE servers
• 2-4 GB memory per CPU core

High availability - points to consider

• Hardware/Software load balancers
• Cluster/Mirror of Sql server
• Service applications associated to the app
• Search target, dedicated/distributed

Sql Server memory recommendation

• Small - 8 GB
• Medium - 16 GB
• To handle up to  2TB of data - 32 GB
• To handle data from 2TB to 5TB - 64 GB
• To handle more than 5 TB - Go for new instance of Sql server

As there is no perfect tool (yes there is a tool called HP Sizer is available)to tell you what kind of Farm is required for your purpose, it requires a careful assessment of all these points and the requirement is necessary to arrive at a conclusion. It is always recommended to do oversizing rather than undersizing(Want to know your servers are over sized/undersized, use performance monitor counters to gather data and analyse).

I guess most of the points were taken from this Technet Article. Nevertheless for those who need full set of information please go through that link. I don't find the deck presented from Tech Ed site, will provide a link when it's available.