storm-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "P. Taylor Goetz" <ptgo...@gmail.com>
Subject Re: Storm (Core) versus Storm (with Trident)
Date Wed, 25 Jan 2017 22:58:11 GMT
If you could provide additional information about your environment (as much as possible) that
would help us answer your questions.

-Taylor

> On Jan 25, 2017, at 5:35 PM, Thomas Cristanis <thomascristanis@gmail.com> wrote:
> 
> Thank you, P. Taylor.
> I asked this question because I'm making a basic benchmarks and I have the opposite result.
The Storm Core (Kafka) is about 2x and tries more than Trident (Kafka).
> I still do not understand what's happening.
> 
> 
> --
> Thomas Cristanis
> 
> 2017-01-25 19:21 GMT-03:00 P. Taylor Goetz <ptgoetz@gmail.com>:
>> It comes down to a tradeoff between throughput and latency (as well as how and what
you do in your topologies), and tuning parameters.
>> 
>> In the benchmarking I’ve done, Storm Core has lower latency and lower throughput,
Trident has higher throughput (~2x) and higher latency (~2-3x).
>> 
>> But again, it really depends on what other systems you’re interacting with. One
slow external system could easily negate any performance difference between the two.
>> 
>> -Taylor
>> 
>> 
>> > On Jan 25, 2017, at 4:03 PM, Thomas Cristanis <thomascristanis@gmail.com>
wrote:
>> >
>> > Has anyone done any benchmarks comparing the performance of the Storm against
the Apache Storm (Core) in relation to Apache Storm (with Trident) ?
>> > If so, what were the results?
>> >
>> > --
>> > Thomas Cristanis
>> 
> 

Mime
View raw message