Microsoft Office Tutorials and References
In Depth Information
How many sessions will be active at peak?
What types of operations are users of Excel Services performing? This includes user roles (per-
missions), actions being performed (such as PivotTable drill operations), and interactivity.
How much throughput is needed from Excel Services? What are the response time expectations
for Excel Services?
Addressing the capacity requirements of the system may require that Excel Services be scaled out, scaled
up, or possibly both. The direction your system takes will come down to the answers to the previously
noted capacity questions. The following sections discuss how the answers affect capacity.
For Excel Services, to scale out implies adding additional servers to the farm to benefit Excel Services. You
can either add front-end servers where the Excel Services front-end web components are located, or add
application servers where the ECS components are located.
In Excel Services, to scale up commonly applies to increasing system memory, but it could also apply to
increasing available processors. You can scale up an ECS by replacing an existing four-processor server
with an eight-processor server.
The administrative controls provided by Excel Services, as well as the best-practice implementation directed
at workbook authors and consumers, can also affect Excel Services capacity. Ask yourself these questions:
Which security model will be used for accessing external data? How many different sets of
credentials will be used when querying results from a data source? Will the use of trusted data
connections be enforced? Will ECS data cache lifetimes be long or short?
Will SharePoint roles be used extensively, or will most users be operating with elevated
privileges? How much interactivity will the typical user perform when using the EWA? Will EWA
toolbars be customized and tailored to expose only relevant and necessary functionality?
How many dashboards will be used with the EWA? How many sessions will each dashboard
generate? How long will a typical dashboard session last? How often will sessions be private
instead of shared? How many concurrent active sessions will each user be allowed to create?
Will authors create calculation intensive workbooks? Will UDFs be used? How often will volatile
functions be used, and will the volatile function cache lifetime setting be long or short?
Will the Excel Services Web Service be available to a large part of your workforce? Will code
authors consistently close sessions and be responsible users of the API?
One of the main areas you need to focus on when capacity planning is the amount of memory consumed
by the ECS process. The second area is CPU usage. The ECS is a memory-intensive service. The ECS can
also be CPU-intensive, but that is typically isolated to performing heavy calculation operations. If you
are not accustomed to seeing a process size grow beyond 1GB, then you may be a bit alarmed when you
see the ECS process size after servicing requests for a while. In a 64-bit environment, the process size is
allowed to grow even more, provided system resources are available to the process.