End Point

News

Welcome to End Point's blog

Ongoing observations by End Point people.

RailsConf 2008 Report

End Point's Sean Schofield recently returned from a trip to RailsConf 2008. RailsConf is the annual gathering of Rails developers which was held for a second year in a row in Portland, Oregon. There is also now a European version of the conference which is held during the fall. The conference consisted of one day of tutorials, followed by three days of sessions and keynotes.

Attendance was extremely heavy this year (over 1,800 people) which caused some initial crowding issues with the sessions. Fortunately, these issues were resolved by the second day and the conference organizers even managed to schedule a repeat of the first day's talks for those who were initially shut out.

Several notable Rails personalities were speaking at the conference this year, including David Heinemeier Hansson, the creator of Rails. Other speakers of note included Obie Fernandez (author of The Rails Way), Joel Spolsky and David Chelimsky (of RSpec fame).

Scott Chacon gave an interesting talk entitled Using Git to Manage and Deploy Rails Apps. The Git distributed source code management system has been taking the Rails world by storm this year. Two factors have contributed to this explosion in popularity. The first is the advent of the popular Github service which makes it very simple to contribute patches to an open source Rails application. The second was the announcement that the RoR project would be dropping Subversion in favor of Git.

Another very popular talk was given by Yehuda Katz, who spoke about various aspects of the up and coming Merb framework. There was a steady buzz regarding Merb that was present for the entire conference. A significant portion of the Rails developers that Sean talked with were actively considering the use of Merb in their next project (if they hadn't used it on a project already).

The single hottest topic at RailsConf 2008 was Phusion Passenger aka "mod_rails." Phusion Passenger is a replacement for the commonly used Mongrel cluster approach for deploying Rails applications. It's intended to allow for simple deployment using the industry standard Apache web server. Everyone was talking about this hot new technology at the conference, including several of the exhibitors.

A veteran of several technology conferences, Sean was impressed with the overall quality of the presentations. He considered the technical content to be highly informative and appreciated the professional quality of the speakers. Please see the conference site for a complete list of the talks as well as presentation slides.

PGCon 2008 Report

End Point's Greg Sabino Mullane recently returned from PGCon 2008, the annual conference for the PostgreSQL database project. The conference was held in Ottawa, Canada, and is a mix of Postgres developers, companies who are using Postgres, students, and everyone else involved in the vibrant Postgres community.

Greg presented a talk on Bucardo, the multi-master and master-slave replication system for Postgres. He explained the strengths and weaknesses of Bucardo, its typical use cases, and described in detail how it works. He detailed the innovative use of "hooks", which allow custom code to be fired at any point throughout the replication process. The hooks are also the method of doing conflict resolution and exception handling, two important factors for multi-master replication. He also discussed the use of DBIx::Safe to pass restricted database handles to the custom code, as well as future directions for the Bucardo project. The talk even ended on time and left time for questions. The slides are available on the PGCon 2008 site.

The other talk Greg gave was a "lightning talk" on DBIx::Cache, a query caching system for Postgres built on top of DBI and DBD::Pg. (Lightning talks are a collection of small five minute talks by different people, highlighting things they are currently working on.) There was a high level of interest in DBIx::Cache, which is still under development but should have its first released version within a few weeks.

There were many other interesting talks at the conference, many focusing on ways that companies and developers are pushing Postgres to new heights of performance and scalability. Yahoo! announced their new petabyte-sized database built on top of Postgres, and NTT presented an innovative way of doing synchronous log shipping for extremely fast failover capability. Andrew Dunstan described a replacement to the existing listen/notify system that he was working on, which will not only be faster and more reliable than the current system, but will allow the use of "payload" messages as well. Greg added support for payloads to DBD::Pg the next week, so we're ready when you are, Andrew!

Pavel Deolasse of EnterpriseDB gave a great talk about Heap-Only Tuples (HOT), a clever new feature in Postgres 8.3 that significantly improves performance by only updating indexes when absolutely necessary, and many other optimizations. In other words, if you have a table with columns a and b, with an index on column a, updates that only affect column b will not change the index on a at all.

There were many other talks, covering a wide range of things, from PostGIS (geographical extensions) to upgrading Postgres on the fly, to satellite data processing by NASA. The full schedule can be seen here. Just as valuable as the talks were the discussions among attendees over dinner and between sessions, about current problems, brainstorming future features, comparing war stories and victories, and catching up with people not seen for a year or more.