Concurrent requests to DRPC

[Douglas Alan <do...@alum.mit.edu>]

Hi. We want to make a Storm topology than handles DRPC requests, and we
want the requests to be handled concurrently. I.e., if a DRPC request comes
into the topology at time T0 that takes a long time to complete (let's say
100 seconds), we don't want this request to block another request that
comes in at T0 + 1 sec and will only take 1 second to complete. I.e, we'd
like to get the result for second request at time T0 + 2 sec  or T0 + 3 sec
or so, and not at time T0 + 101 sec.

Does this behavior happen automatically? Or would we have to do something
special to make Storm behave this way? Or is it just not the way that Storm
works? I haven't found anything about this in the online Storm
documentation or in any books.

I looked through the user@storm.apache.org archives looking for an answer
to this question before posting. I found this question posted a few months
ago:

>From "JianJian, Bu" <ji...@amazon.com>
Subject Issues about concurrent request to Trident DRPC
Date Mon, 29 Dec 2014 07:08:10 GMT
Dear All,

We’re encountering some issue when use trident DRPC --  It seems that
trident DRPC request is executed sequentially with later request blocked by
previous one no matter how the parallelism parameter would be set.

Our usage is :  we setup a client faced web services  which is hosted by a
group of hosts. The web service receive requests from our client and call
storm DRPC to query the data and send back the result to the Client.

Now the issue is some of the request will take much time to be finished by
DRPC.  So these requests will block other clients from getting the result.
  I can see many introductions in storm wiki about how drpc works but there
is little threads about how drpc handles multiple requests. There is only
one sentence about the concurrent reuqests in this wiki
https://storm.apache.org/documentation/Distributed-RPC.html “KeyedFairBolt
for weaving the processing of multiple requests at the same time” .  But it
has little help. There may be some ideas like start multiple similar
topology in the
same cluster but I think that would have other issue like balance the
load.  Does anyone ever have the similar issue and how did that get solved?

Unfortunately, there are no responses in the archive to this previous
posting.

Is Storm DRPC not widely used?

|>oug