Having looked at the Kudu paper & code yesterday, it's still early days. It might not be worth spending too much time on an evaluation just yet. Quoting from the paper: "In the current beta release of Kudu, this consistency support is not yet fully implemented. However, this paper describes the architecture and design of the system, despite the presence of some known consistency-related bugs."

Generally, I'm glad about this very recent trend of moving back to C++, first with http://www.scylladb.com/ for Cassandra, and now Kudu for HBase. We are spending a good amount of time working around JVM scaling & latency issues, so I'm looking forward to see these projects mature into serious alternatives.

Gabriel

On Tue, Sep 29, 2015 at 12:19 PM, Dan Andreescu <dandreescu@wikimedia.org> wrote:
Thanks Toby,

If anyone's really passionate about discussing the tons of big data tools available, this would be the place: https://wikitech.wikimedia.org/wiki/Analytics/DataStore/Evaluation#Candidates (just added Kudu)

On Tue, Sep 29, 2015 at 3:10 PM, Toby Negrin <tnegrin@wikimedia.org> wrote:
From the intertubes:

@tlipcon: Super excited to finally talk about what I've been working on the last 3 years: Kudu! http://t.co/1W4sqFBcyH http://t.co/1mZCwgdOO5

Might be useful for the media wiki tables. 

-Toby

_______________________________________________
Analytics mailing list
Analytics@lists.wikimedia.org
https://lists.wikimedia.org/mailman/listinfo/analytics



_______________________________________________
Analytics mailing list
Analytics@lists.wikimedia.org
https://lists.wikimedia.org/mailman/listinfo/analytics




--
Gabriel Wicke
Principal Engineer, Wikimedia Foundation