Got your attention? Good!
Brian O’Neill details in A Big Data Trifecta: Storm, Kafka and Cassandra an architecture that was fast enough to choke the Cassandra Bolt component. (And also details how to fix that problem.)
Based on the exchange of tuples. Writing at 5,000 writes per second on a laptop.
More details to follow but I think you can get enough from the post to start experimenting on your own.
I first saw this at: Alex Popesu’s myNoSQL under A Big Data Triefecta: Storm, Kafka and Cassandra.
Interesting post. Also interesting to see the reference to Taylor Goetz at the bottom. He forked Duke on Github, so I recognized the name.
Comment by larsga@garshol.priv.no — August 8, 2012 @ 1:10 am
Thanks for the note!
I catch myself thinking that some problems are recurring and quite fundamental. Or perhaps it is that all solutions are partial and incomplete. 😉
Comment by Patrick Durusau — August 8, 2012 @ 2:25 pm
[…] More details on Kafka (see Choking Cassandra Bolt). […]
Pingback by Building LinkedIn’s Real-time Activity Data Pipeline « Another Word For It — August 16, 2012 @ 1:21 pm