Microsoft SharePoint(TM) Portal Server 2001 Resource Kit (Examples & Explanations Series)
Selecting an appropriate server platform is an extremely important component of a successful portal deployment. The key hardware requirements are the CPU speed, amount of RAM, and hard disk space:
- Sufficient CPU resources allow SharePoint Portal Server to provide an excellent user experience to large numbers of users during peak usage periods.
- If there are insufficient CPU resources, users experience unacceptable server response times during peak usage periods.
- Additional RAM and hard disk space allow SharePoint Portal Server to provide improved user performance.
- If there is insufficient RAM, users experience unacceptable server response times regardless of the number of active users.
- If there is insufficient hard disk space, users are not able to search for or save additional documents.
An important step in determining your server hardware requirements is establishing clear performance and scalability requirements. Unfortunately, it is frequently difficult to establish clear and detailed performance and scalability requirements for portal deployments. Most organizations find it difficult to predict the level or type of use that the site receives. To complicate matters, this level of use frequently changes and grows over time.
Most organizations can estimate accurately the following important deployment metrics:
- Number of users. This is the total number of users who may have access to the site.
- Percent of active users per day. This percentage represents the total number of users who might use the dashboard site during any particular day. Typically, this figure ranges from 10 percent to 100 percent. This number is frequently overestimated and is commonly around 30 percent.
- Number of operations per active user per day. This is the number of operations that a typical user does from the dashboard site during a typical day. An operation is an action such as viewing the home page, searching, retrieving documents, etc. This number usually ranges from 1 to 10. It is frequently possible to estimate the number of operations by analyzing the Web server log of an existing portal deployment, if one exists. Note, however, that when analyzing the Web server log it is only necessary to consider page views, not site hits. Site hits are frequently significantly higher than the number of page views.
- Number of hours per day. This is the number of hours during which most activity occurs. This number typically ranges between 10 and 24 hours.
- Peak factor. This is an approximate number that estimates the extent to which the peak throughput for the dashboard site exceeds the average throughput. This number typically ranges from 1 to 5.
You can use the quantitative descriptions of a portal deployment to estimate the required peak throughput. The following formula yields the peak throughput in operations per second.
Peak Throughput Formula
Number of users | X | percent of active users per day | X | number of operations per active user per day | X | peak factor |
360,000 | X | number of hours per day | ||||
The number 360,000 is determined by:
100 (for percent conversion) X 60 (number of minutes in an hour) X 60 (number of seconds in a minute)
SharePoint Portal Server uses Hypertext Transfer Protocol (HTTP) for all communication between the client and the server. The HTTP protocol is a connectionless protocol. Therefore, it is not possible to identify the number of concurrent users. The most important measurement of server throughput is the operations per second. The following examples illustrate applying the formula to a sample deployment for three sites.
Deploying a Small Group Site
The following is an example of how you can use the preceding formula to determine the requirements for a product group portal for 200 people. The product group contributes more than 90 percent of the site traffic. Although there might be thousands of other users who occasionally use the site, the number is insignificant compared to the product group usage.
For such a deployment, the following characteristics are reasonable.
| Small Group Site Deployment | |
|---|---|
Number of users | 200 |
Percent of active users per day | 90 |
Number of operations per active user per day | 10 |
Number of hours per day | 12 |
Peak factor | 5 |
These estimates yield a predicted peak throughput of 0.21 operations per second.
Predicted Peak Throughput
200 | X | 90 | X | 10 | X | 5 | = | 0.208 |
360,000 | X | 12 | ||||||
You could successfully deploy such a site with a server such as a 500 megahertz (MHz) Pentium III with 512 megabytes (MB) of RAM.
Deploying a Large General Site
Another common deployment scenario is a divisional or enterprise portal for approximately 2,000 users. Such a site usually differs from the small group site as follows:
- A more diverse set of workers uses the site.
- The user community is frequently not as focused on document publishing or document collaboration.
In this scenario, there is usually a broad mix of user needs. The average user is typically a less-focused user who might access the site to find a document or to read the morning news or announcements.
The following characteristics for such a site are typical.
| Large General Site Deployment | |
|---|---|
Number of users | 2,000 |
Percent of active users per day | 90 |
Number of operations per active user per day | 5 |
Number of hours per day | 12 |
Peak factor | 5 |
These estimates yield a predicted peak throughput of 1.0 operations per second. Such a site would probably require a server such as a quad processor 500 MHz Pentium III with 1 gigabyte (GB) of RAM.
Deploying a Large Search Site
One final deployment example is the corporate search portal. In this scenario:
- The organization uses the portal predominantly to search over content that was crawled from other sources within the organization.
- The organization does not manage a large number of documents from the site.
- Users are more infrequent than users of the large general site.
- Usage spreads out more during the day due to time zone distribution.
The following are common characteristics for such a deployment.
| Large Search Site Deployment | |
|---|---|
Number of users | 10,000 |
Percent of active users per day | 50 |
Number of operations per active user per day | 5 |
Number of hours per day | 16 |
Peak factor | 3 |
These estimates yield a predicted peak throughput of 1.3 operations per second. Such a site would probably require a server such as a quad processor 700 MHz Pentium III with 2 GB of RAM.
A large search site that includes more than 100,000 documents in the index benefits from using a server dedicated to crawling content to improve performance for users of the server dedicated to searching.
Summary of Requirements
The following table presents a summary of the operations per second and the recommended hardware requirements for the three configurations discussed previously.
Deployment Configuration Comparison
| Small Group Site | Large General Site | Large Search Site |
|---|---|---|---|
Number of users | 200 | 2,000 | 10,000 |
Percent of active users per day | 90 | 90 | 50 |
Number of operations per active user per day | 10 | 5 | 5 |
Number of hours per day | 12 | 12 | 16 |
Peak factor | 5 | 5 | 3 |
Operations per second | .21 | 1.0 | 1.3 |
Recommended processor | 500 MHz Pentium III | Quad processor 500 MHz Pentium III | Quad processor 700 MHz Pentium III |
Recommended RAM | 512 MB | 1 GB | 2 GB |