Hello,
Regarding Starfish Profiler.
I am looking into the job profile created by starfish.
We have average timing for map and reduce in the created job profile.
I have one map and one reduce in my job. Reduce is configured to start after completion of map.
The timing of map + reduce (including subphases) should be equal to the total job response time.
However, when I match this time with the job response time from Hadoop History logs, it is always less than the time available from Hadoop History logs.
Can you suggest how can I co relate both timings.
As an example, I am attaching two files:
job profile from starfish and
hadoop history log of the same.
There was only one wave of map and reduce and reduce started after finish of map.
Thanks in advance,
---------- Forwarded message ----------
From:
Amit Sangroya <sangro...@gmail.com>Date: Wed, Mar 11, 2015 at 8:31 PM
Subject: Re: Starfish with TPCH
To:
har...@cs.duke.eduCc: Shivnath Babu <
shiv...@cs.duke.edu>, Herodotos Herodotou <
herodotos...@cut.ac.cy>
Hello Herodotos,
Regarding Starfish Profiler.
I always notice that timing values of (map + reduce) in job profile is always less than the job response time from Hadoop history logs. The job response time should match to the total of timing values in the job profile. Am I missing something?
Thanks in advance,