Generally yes, the best case for output collector is passing a reference
through some queues.  However, it's harder to reason about the performance
of a larger topology, and (assuming you use reliable messaging) your entire
topology can be held up by one poorly performing bolt.
On Nov 5, 2015 7:31 AM, "Crina Arsenie" <[email protected]> wrote:

> Hello,
>
> I'm interested in this topic also. Thank you for your answer.
> I didn't knew about Flux, maybe it could do the job for my case, i'll take
> a lot at it.
> I have a also question about performance, I assume that passing through
> the output collector is faster than Kaka, what do you think ?
>
> Thank you,
>
> Crina
>
> 2015-11-05 12:36 GMT+01:00 Nathan Leung <[email protected]>:
>
>> It's not possible to combine several topologies into one, but it should
>> be possible to write different tuple sinks such that you can configure each
>> bolt to write to either the output collector or Kafka. Then it's just a
>> matter of wiring and configuring your bolts differently.
>>
>> You can use something like flux (
>> http://storm.apache.org/documentation/flux.html) to change how your
>> bolts are wired without having to rebuild your jar file every time.
>> On Nov 5, 2015 4:17 AM, "Irina Alles" <[email protected]> wrote:
>>
>>> Hello,
>>>
>>>
>>>
>>> We are currently studying if storm would be appropriate as a part of our
>>> monitoring system.
>>>
>>> We are measuring sensor data, we need to apply different transformation
>>> steps and store it somewhere or send it further.
>>>
>>> This seems to be a basic use case for storm so far, let’s call this
>>> setup topology A.
>>>
>>>
>>> We would like to be able to add measured sensors and their
>>> transformation steps (topology B) dynamically without requiring any system
>>> downtime. These dynamic additions could happen frequently. It will happen
>>> that bolts of topology B will require the output of certain bolts in A
>>>
>>> Is there a best practice to manage this in storm?
>>>
>>>
>>>
>>> After a certain runtime of the system we will have to manage several
>>> topologies and we need to assure the communication between them.
>>>
>>> We thought about using Kafka for the communication between topologies,
>>> but with the growing number of topologies this might not be the best
>>> approach I suppose (‘best’ means in this case: easy to handle, avoiding
>>> message overhead).
>>>
>>> Maybe it would be better to group some topologies to create a greater
>>> one? How would one do this in storm (I’ve read about the swap feature, but
>>> it doesn’t seem to be available yet)?
>>>
>>> Is there a better approach?
>>>
>>>
>>>
>>> Thanks!
>>>
>>> Irina
>>>
>>
>
>

Reply via email to