For the most part, JavaOne takes place between Monday morning and Wednesday afternoon. Sure, there's a keynote or two on the Sunday evening, but at that point int time, many attendees are still flying in and checking into their hotels. There are also a sessions on the Thursday, but they get minimal coverage, as the exhibit floor is no longer open, and people are either flying home, or too hung over from the OracleWorld concert that took place the night before at San Francisco's Treasure Island. Fortunately for us though, Shaun Abram stuck around, and gave us this update about what was happening on the Thursday, while most of us were flying home, and the staff at the Hilton were cleaning up the conference rooms.


I attended the final day of sessions at JavaOne on Thursday, focusing on 'Big Data' and NoSQL talks.

"The Cassandra Distributed Database" talk from Jonathan Ellis (formerly a Systems Architect at Rackspace Hosting, now at Riptano, a Cassandra support and services company) gave an introduction to this open source, NoSQL (i.e. non-relational) database, which is now an Apache top level project. Jonathan talked about the issues many companies face with database scalability and common solutions, including horizontal and vertical scaling, introducing a cache, and replication, before examining what Cassandra brings to the table, such as scalability, performance, reliability (no single point of failure) and monitoring capabilities. It does all this via a loosely schema-less design which while not quite a key/value store (it does provide rows/columns), does always require access via a primary key. It seems Cassandra is becoming a very popular NoSQL solution, with large implementations at Twitter, Facebook and Digg.

Shevek M from Karmaspehere gave a talk on "Practical Big Data Processing with MapReduce and Hadoop". The talk had much more of a focus on MapReduce than Hadoop, but gave a very interesting history of computing along the way (including the “Enigma? machine and the idiosyncrasies of Cray computers!). As well as covering the basics of MapReduce (see Google’s definitive white paper for more), Shevel also talked about some of the practical implementation issues you are likely to run into.

The final NoSQL talk of the day (and of the conference!) was from Rod Cope of OpenLogic, talking about his "Top 10 Lessons Learned from Deploying Hadoop in a Private Cloud". Rod shared with us his hard earned lessons from implementing a large Hadoop cluster using HBase (which is like BigTable for Hadoop) and Solr (a search server based on Lucene). While Rod and company developed a very (very!) scalable and reliable solution, it certainly sounded like the whole process was a huge undertaking with many hurdles along the way.

I also managed to attend the much more light hearted, but none-the-less useful talk "97 Things Every Programmer Should Know" by Kevlin Henney and Kirk Pepperdine, authors of the O'Reilly book by the same name. This was a hugely entertaining talk containing pearls of advise on how to survive life as a humble programmer but while the talk was riddled with humor and amusing anecdotes, that didn't take away from the invaluable (and sometimes profound) advise given by the speakers and authors.

Overall, a very useful final day at JavaOne. The scattered layout of the sessions and the fact that so many classes seemed to be booked out so far in advance, certainly made it challenging to be where you wanted to be (and was I the only one who found the online schedule builder really tricky to use?!), but hey, it was JavaOne! Worth the effort...