Whenever we decided to deploy SharePoint portal server the first thing that came to the mind is about the specification of systems that is required for the successful deployment of the portal. In this post I will try to explain the way to estimate the requirements for the system.
SharePoint Portal server deployment differs from organization to organization. In my case where I worked SharePoint deployment was mostly used as General purpose e.g. Portal Site for organization then departmental areas and sites, then project team sites and document storage, meeting workspaces. Some organizations use it for Searching in my current case we are using it to search across 200 GB of data available on file shares. There are also few scenarios where companies use them for teams and collaboration. Whatever the design is we need to analyze the following
Size of Documents/artifacts that will stored in SharePoint server sites, Document libraries, Lists, etc.
Estimate any external content that is required to be crawled or indexed
Users working style and actions that they perform
Each knowledge worker normally uses Office system as client and connects to SharePoint server as server and collect data from different places during his working timings. Each user action is called transaction. Depending on the size and configuration of the system we need to calculate the frequency of transactions. The transactions arrivale rate is called throughput rate. When a user visit the portal and uses the content of various areas and sites, click links, add tasks to tasks page, add or edit items in document libraries; all these action can be combined to calculate the throughput rate. Each of this action is given a number by a simple and complex actions matrix. The following points can be used to create this kind of matrix.
Number of Users
Percent of active Users Per day (visitor per day that comes to portal Aprox 20 to 50 percent.)
Number of actions per active user per day (Access to different areas of the portal e.g. home page, divisional areas, my site, project sites, document libraries, etc.)
Hours per Day (number of working hours. 8 in General (PK)
Peak factor: a number that represents the how much ratio of usage exceeds peak usage of the portal
The formula to estimate the Peak Through put will be
Throughput = NU x PAUPD x NOPAU x PF x 3,60,000 x NHD
|NU||Number of Users|
|PAUPD||Percent of Active Users per Day|
|NOPAU||Number of Operations per Active User|
|PF||Peak Factor (1 to 5)|
|NHD||Number of Hours Per day|
|3,60,000||100 % * 60 Minutes per hour * 60 Second Per Minute|
As portal site will contain many different sections e.g.
Main Organizational Portal site
We need to calculate the breakdown of operation for the above listed sections as well. Below I will try to use a sample of to calculate a throughput. As we know that site usage depends on the users. It is observed that almost 60 percent of the portal users visits the site (Here in PK) for general purpose. Their main goal is to add or view a link, view some information on their personal sites and etc. The rest of 40 percent are those who perform complex actions like adding documents, creating workspaces. I have thought of an average company that has 400 users and other details are given below.
|Common Actions per day||36 * (weight = 1.0) = 36|
|Complex Actions per day||20 * (weight = 3.0) = 60|
|NOPAU (common + complex)||96|
The throughput can be calculated now which should be as follows
400 users x 80 % usage x 96 Operations x 4 peak = 17.07 Ops/sec
AS these characteristics reveals the predicted throughput requirement of 17.07 Operations per second. To run just a medium server farm with two load balancing front-end web servers, an indexing servers, and an server running sql server components would be needed. A rough estimate for performance is each 1 GHz of processing power in the front end servers can support 9 to 10 operations per second and each server should be approximately 2x 2-GHz Pentium 4 Server with 2GB of Ram. For our example we can have a small and medium server farm with one front-end web server, 1 search and indexing server and 1 server with SQL Server components installed. Each machine in the server should be at least 3.4GHz Pentium 4 Server with 4 GB or RAM. This could be close to our example. Of course if organization afford hardware then the example given above would be perfect.
The numbers provided above are general guidelines and approximations, as result will vary based on the exact hardware you are running. Do run the planning test on your servers to ensure that they will be able to perform adequately specially for large server farm environments.
In the above post I have tried to show an example on how to estimate the system requirements for SharePoint portal server deployment. Planning the deployment is vital as most of the implementers skip that part which result in failed deployment of the server.