Legacy ID:KA293069
The 'Estimating disk space for history data' in the BMC Performance Assurance Getting Started guide (http://documents.bmc.com/supportu/documents/89/18/68918/68918.pdf) is a good starting point and is applicable to UDR capacity planning data as well. What changes is the 'nIntervals' value. In both cases it is the summarization interval. For UDR capacity planning data 'nIntervals' will be the spill rate/summarization interval selected (by default 15 minutes - so '96' for a day). That section doesn't talk about the collection of configuration groups so it will somewhat understate the overall requirements. 'Disk Configuration' will generally be about 40% of the size of 'Disk Statistics' and 'Logical Volume Configuration' will be about 100 bytes per logical volume per summarization interval.
The problem with calculating disk space requirements is that the data size is directly related to the number of 'objects' being collected so its usually a lot of work to get an accurate count of those things in order to calculate the size (a lot harder than just collecting some data and seeing how big it is since the data size collected on a machine typically doesn't change much from day to day).
For the most part there are 3 or 4 metric groups that will represent 80% or more of the total size of the day's UDR data. Usually 'Process Statistics' is the largest group by far. Then, on machines with a large I/O subsystem you'll see a large 'Disk Statistics' group. You may also see a large 'Disk Configuration' group (although it will generally be less than half the size of Disk Statistics) and if there are a very large number of logical Volumes a relatively large 'Logical Volume Configuration' group. Beyond that most of the other groups are generally relatively small.
Q: How would various data summarization intervals (spill intervals) impact the amount of data collected?
Spills of 1 minute, 5 minute, 10 minute, 15 minute, and 60 minutes are all possible within the product. For the most part the cost of faster spills is a little bit of additional CPU at the Perform Agent level and 80-100% more disk space utilization each time you double the summarization rate.
So, for example, if your current 15 minute summarization interval results in 15 MB of data over a 24 hour period it would look something like this:
1 second -- Not possible
1 minute -- 240 - 300 MB
5 minutes -- 48 - 60 MB
10 minutes -- 46 - 45 MB
15 minutes -- 15 MB
60 minutes -- 4 - 8 MB
Most of the groups are just a strict multiple of whatever their current size it. The big variable is Process Statistics. A process record is written into the UDR data once per spill or if the process is around for less than 1 spill at the end of its life. So, if you have a lot of short lived processes then your Process Statistics group won't actually grow by a huge percentage as you summarize more frequently (since there will still just be one record written to the database). If you have a lot of long lived processes then your Process Statistics UDR data will just grow by the ratio of the new summarization interval to the old summarization interval (since the process will still be written once per summarization interval but there will be more intervals.
If you aren't already collecting UDR data on the machines then it more difficult to determine how much space will be required because it is based upon a lot of variables (how many disks, how many file systems, how many processes per interval, and so on).
Related Products:
- BMC Performance Assurance for Servers
- BMC Performance Analyzer for Servers
- BMC Performance Analysis for Servers