I was working on a OMS project a few days ago and came across a new feature called Near-Real Time (NRT) Performance Data Collection. I have been anticipating the release of this feature, so I was very excited to see that it had been released! Let’s take a quick look….
First, let’s enable the NRT Performance Data Collection feature:
- Navigate to Settings–>LOGS
- On the right side of your screen you will see a new section called “Performance Counters”. There are several options listed which you can add by simply clicking the Add button, in addition to a search bar in which you can enter any object or counter to find additional options. When you start typing you will see auto suggestions which is very helpful when trying to find specific counters. Very cool!
- Notice the Interval setting. You are able to set the collection interval to as low as 10 seconds for each counter. WOW. Those that have experience using OpsMgr know that high frequency performance collection has been a major blocker for companies with large agent counts who want to use OpsMgr for root cause analysis. The database simply can’t handle the load. OMS NRT Performance Data Collection provides us a way to collect this high frequency performance data without writing to the OpsMgr databases at all!
Let’s take a look at the functionality….
- On the Overview page select the Log Search blade.
- In the query search bar enter the following: Type=Perf
- Notice all of the fields listed on the left side of the screen (see below). These fields can be used to further filter your queries to display more granular searches.
- Let’s filter our query to just show data for the Disk Transfers/sec counter.
- See the image below for detailed explanations for the output. NOTE: There are two perspective views for this search. The first one (Logs) will return the 30 minute aggregate values for all the performance counters, and the second one (Metrics) will return the visualization of the raw results for each of the returned counters. Currently, the raw data is stored for 14 days. The 30 min aggregates will be stored based on your data plan with OMS.
From the official TechNet blog here:
If you like to start using this feature, but don’t know the estimated usage for performance counters, this example may help you.
For a particular computer, a given counter instance (e.g., Processor(_Total)\% Processor Time) with 10 second sample interval will send ~1MB per day (~1MB/day/counter instance). You can multiply this number by the number of computers that you have to get an estimate usage per counter instance across all your computers.