Cassandra Summit Recap
I want to start this post with a huge thanks to our own Lynn Bender for all the hard work putting the first Cassandra Summit together and of course to Rackspace for co-sponsoring this event with us. With over 200 attendees from as far away as Tokyo and Zurich I think we can call it a success.
The best place to start off the actual recap is with Jonathan Ellis’s keynote speech. It is informative as set of slides by itself, but the delivery really set the tone for the day as a good keynote should. Watch this now if you have not already (video will be available soon).
After the keynote, my first stop was Stu Hood’s “Partners in Crime: Cassandra Analytics and ETL with Hadoop” as it’s been obvious to me that a lot of folks in the Cassandra community are real interested in Hadoop integration. Stu gave a good demo and fielded some tough question for a very knowledgeable audience. Given the amount of interest here, I’m looking forward to some really cool things to come out of Hadoop and Cassandra integration in the very near future.
The next presentation I went to was Gary Dusbabek’s code walk through. Anyone at all interested in how Cassandra works or, even better, in getting involved in development should take a look at this presentation. Keeping to the one hour time constraint given the complexity of a system like Cassandra must not have been easy. Gary did this and managed to cover a lot of ground with what I felt like was just the right amount of detail.
Next up for me was Ben Black’s presentation “Operations and troubleshooting in Cassandra”. This turned out to be like a live version of the #cassandra IRC channel. I don’t mean that dismissively in any way – rather he took some of the most common issues repeatedly encountered and stepped through their causes and solutions clearly and concisely. Judging from the questions and some of the discussion I overheard afterwards, the new users in the audience got a lot out of this.
The lightning presentations wrapped up the day nicely for me. Noah Silas and John Watson from Mahalo provided a good case study of their usage of Cassandra. They also detailed their plans for expansion of their cluster to another data center which I think a lot of people found useful.
David Strauss presented Cassandra as the core of a message queuing system. I’ll describe this now only as intriguing as I need to look at it in more detail to do an explanation justice.
I came up next and was able to show off some of the changes we’ve been working on in Hector. If you have not heard about it yet, our new higher level API completely encapsulates Thrift while providing strong typing (apparently Cliff Moon was talking about this type of approach in the other room at the same time – I’m sorry I missed it). The biggest win is that you get all this on the solid foundation of pooling, monitoring and failover capabilities that made Hector popular in the first place.
Paul Querna and Dan Di Spaltra (“the Cloudkick guys”) finished up the day with a good case study of their cluster. This is the second time I’ve seen these guys talk about how they use Cassandra and not only did I found their presentation informative (particularly their ability to do what they do with only 8 machines), but I think their openness and willingness to share their experiences is a credit to the community we are all working hard to build.
Thanks again to all the attendees and speakers who made this such a great event.
Note: If you are looking for details of the other talks which I did not attend, Bryan Duxbury has a good summary (which is ironic for me because I was looking forward to meeting him, but never seemed to track him down).