Re: how to calculate the storm's throughput ?

731 views
Skip to first unread message

Will Koffel

unread,
Oct 9, 2012, 5:16:29 PM10/9/12
to storm...@googlegroups.com
If I understand your question correctly, Chrono, you are wondering how to get a number for the throughput of your storm topology.

I believe that "throughput" is going to be dependent on what you're trying to measure exactly.  If you want to know the end-to-end acknowledged tuple throughput, you can divide the "ack" value by the time.  For example, in the Storm UI (or Thrift interface), there's a 10 minute window listed.  If you acknowledge 25,200 tuples during that time period, then your throughput rate for the whole topology was 42/sec.

If you are looking at all the work the whole system is doing, then you probably want "emitted / time".  If you have emitted = 720,000 in 10 minutes, then the storm topology is processing (and perhaps acknowledging, depending on your topology) at a rate of 1200 tuples/sec.

-Will


On Thursday, September 27, 2012 1:15:20 AM UTC-4, Chrono wrote:
Hello.

I'm prepared to try Distributed Processing system - performnace Testing.
My share is STORM.
Yet the analysis. Some are confused by the monitoring.
Q1. can i see the throughput by storm cluster? ( only storm UI and storm api )
Q2. if can't see.. how to monitoring the throughput by storm cluster?
Q3. how to calculation throughput by them( known item < latency, emmited, transfered >, unknown item < turple pass > )?
ps. i known metric , task start, task end, emmited call count, trasfered call= count... but... Much is not enough ... Experience.Desperately try to ask.
ps2. English is not enough ..I'm really sorry.

xmdxcxz

unread,
Oct 10, 2012, 10:54:14 PM10/10/12
to storm...@googlegroups.com
HI,  the method you mentioned above can  calculate the number of tuples, but each tuple has different size. So if we want to calculate the throughput , such as how many bytes the spout received every second. How to hit it?

在 2012年10月10日星期三UTC+8上午5时16分29秒,Will Koffel写道:
Reply all
Reply to author
Forward
0 new messages