Implementing Real-Time Trending Topics in Storm(michael-noll.com)
michael-noll.com
Implementing Real-Time Trending Topics in Storm
http://www.michael-noll.com/blog/2013/01/18/implementing-real-time-trending-topics-in-storm/
5 comments
Michael Noll has some really good introductory articles to Big Data, and I highly recommend his posts for anyone starting in this field.
His post on writing Python streaming programs in Hadoop - http://www.michael-noll.com/tutorials/writing-an-hadoop-mapr... - got me started in this area two years ago, and that's what I'm still doing these days! :)
His post on writing Python streaming programs in Hadoop - http://www.michael-noll.com/tutorials/writing-an-hadoop-mapr... - got me started in this area two years ago, and that's what I'm still doing these days! :)
I read this one a few months back & felt, like most programmers, that it was an awesome article, but I would forget all of it unless I rewrote everything from scratch all by myself :) So here's my rewrite in Scala - its not realtime, but still takes only 4 minutes on my laptop for a typical weekly dataset, so you can put it on a cron schedule & get a trending topics with a 5 min delay. Ofcourse, no storm, no hadoop etc. just a plain old scala program anybody can run on their laptop. http://bit.ly/J82szS
Maybe check out Spark [1], it primarily supports Scala and they have great streaming functionality that seems to beat out Storm and others in terms of performance for some workloads [2].
[1] http://spark.incubator.apache.org/
[2] http://www.cs.berkeley.edu/~matei/papers/2013/sosp_spark_str...
[1] http://spark.incubator.apache.org/
[2] http://www.cs.berkeley.edu/~matei/papers/2013/sosp_spark_str...
Sorry for the shameless trolling, but if you are a fan of storm and see potential in the real-time side of big data, consider applying to work on the Platform team at WebAction.
We're a well-funded company (raised an $11m Series B from a $15billion private equity firm) real-time big data analytics company for the enterprise. We're based in Palo Alto. If interested, feel free to shoot us a message!
[email protected]
We're a well-funded company (raised an $11m Series B from a $15billion private equity firm) real-time big data analytics company for the enterprise. We're based in Palo Alto. If interested, feel free to shoot us a message!
[email protected]
Storm (and Trident) are just fantastic. If you're looking for a streaming complement (or replacement!) for Hadoop, you can't go wrong here.
I did not parallelize the reduce part. I'd need to do some more thinking on how to do that, as I split it into two operators. I think it shows the flavor, though.
Two caveats: One, I have not compiled and tested this, so errors may abound. I was just curious what the "flavor" of it would look like in our system. Two, standard disclaimer of I speak in an unofficial capacity, what I say does not represent IBM or Streams.