topic Re: Splunk optimization : dedup or latest()? in Monitoring Splunk

Splunk optimization : dedup or latest()?

timmalos — Fri, 20 Sep 2013 07:10:26 GMT

Hi
For last 2 months i created a lot of dashboards on Splunk.
In most of my searchs i need to get only the last series of results.
I see i use 2 methods, both work well, but i would know which one is more optimized.
Sometimes I use

search|dedup MyField sortby -_time|table MyField aField bField

and sometimes I use

search|stats latest(aField),latest(bField) by MyField

What is the best to use for short periods/long periods?

Re: Splunk optimization : dedup or latest()?

kristian_kolb — Fri, 20 Sep 2013 07:45:49 GMT

I would think that dedup should be slightly faster, but not by a big margin. Best way to find out is to use the Job Inspector for the searches and look for yourself. It's far more important to limit the initial search, i.e. what comes before the first pipe. Specifying correct index, host, sourcetype and time-limits will have far more effect on the search performance.

However, your dedup syntax looks a bit off, shouldn't it be sortby instead of by?

http://docs.splunk.com/Documentation/Splunk/latest/SearchReference/Dedup
http://docs.splunk.com/Documentation/Splunk/latest/Search/UsingtheSearchJobInspector

Re: Splunk optimization : dedup or latest()?

timmalos — Fri, 20 Sep 2013 09:22:08 GMT

Ure right for dedup, i wrote all from mind so mistakes 🙂

Re: Splunk optimization : dedup or latest()?

sowings — Tue, 24 Sep 2013 13:51:26 GMT

In a distributed environment, stats is likely to be faster, because the indexers can "prestats" before sending their results to the search head. I think that dedup may do a little of this, but the notable distinction is that dedup can't be used with report acceleration searches. I tend to use stats for this purpose.