SSRS 2016 Server capacity Planning - Hardware sizing

Question

I wanted to estimate the required hardware configuration for SSRS 2016 with below user, and reports. SQLServer is used only for ReportServer db and datawarehouse is on another server.

Total Users base: 8000 Concurrent users: 1000 Total Number of Reports: 500 Datawarehouse (Oracle) is on an independent server so no capacity planning required for data query.

Are there any guidelines or approaches to calculating the capacity?

I was considering it as - (Total Concurrent Users * Report data size in mb). + buffer for raw dataset as 25% of ((Total Concurrent Users * Report data size in mb)) + OS Requirement (10% of total RAM)

Example: (1000 * 2mb - report size) = 2000 mb + 500mb (25% buffer for raw dataset) + 5000 mb (OS) so total for SSRS = approximately 3 gb and Total OS 5 gb and grand total = 10 gb approx.

• SQL Server 6 gb for data storage and query. so total RAM required = 10-16gb.

Will this be the correct approach? any suggestions?

Answer 1

Honestly, seems a little complicated of a formula to me, but might be fine, really no way to tell without knowing your data, except to test. But I would think the main factors for Memory are going to be dependent on how much data you think will be concurrently loaded on the server at one time. I wouldn't worry as much about the individual report file size itself, rather I'd focus more on the amount of raw data the report utilizes at a time (e.g. displays or consumed to generate).

For example, a report could theoretically show 1 GB worth of data on a single page. Or even more realistically, if your SSRS reports do a lot of analytical processing of the raw data, then even if the report itself is only showing the aggregated results which is a small amount of data, the raw data it calculated those results off of, in Memory, could be 1 GB alone - for a single report execution. If all 1,000 concurrent users accessed that report at the same time, that would not be a fun time.

Not sure how much data your reports plan to realistically utilize, but it's really going to come down to how much raw data each report consumes and how many of those reports will be concurrently processing data.