Day 1 @ PGCon 2008
By Nirmay26-Oracle on May 23, 2008
On the first day of the main conference here at PGcon, I was awake early with continued excitment of having won the Nintendo Wii and a slight hangover. I got ready early and after finishing some mails headed for breakfast. I was feeling good not only about the Wii but also because of the fun and the geeky mood here at the conference. The last 2 days had been fantastic with the tutorials and the start of the main conference today with a very interesting line up of talks was very exciting.
After the breakfast we arrived at the venue of the conference where registrations were still going on. The opening talk (keynote) by Bruce was good. The main focus for the next years for postgresql is clear. The project will work towards making effective use of multicore/multithreaded systems which are becoming a commodity now. It made total sense to me.
The first session I attended was "postgresql: from a Java Enterprise point of view". Given my long association with Java as a developer I thought this will be interesting to see how fellow java developers think of postgresql. The talk covered the J2EE aspects of entity beans, EJB/QL fairly in detail with changes in advanced versions. I was also hoping to hear how postgresql is perceived in J2EE application developer community, but that was not covered. It was more of PG user - Java developer talk than for the hackers.
I attended Logic and databases session with no expectations but simply thinking what would be the content of this one. I thought it was a very interesting talk for the users/hackers going back to the basics of relating logic to the query language and how a better understanding of this could improve the queries that one writes. Well, I learnt about some good pitfalls and that helped improve my understanding. Thanks Jeff.
then I attended Susanne's talk on "What PG could learn from MySQL". She put together the content based on her experience and with some help from her colleagues both in PG and MySQL commuity. Her analysis was based on various aspects like development model, packaging, User interface, user/customer focus and I agreed with most of the data that she presented, but not necessarily all the conclusions. One thing I liked about the pg hackers during the talk was that they were very receptive about what she had to present and I am sure that will certainly help postgres project.
"Problems with PostgreSQL with multi-core systems with Multi tera-byte data" by Jignesh Shah. To me, this was the best talk so far at the conference. He presented his study with a lot of data on the performance of postgres on multi-core high end systems. The graphs were complicated but he made some very good points about implementaion 0f postgres and also presented his thoughts on how this can be improved. It was a very detailed talk and I must say very well presented to the hacker community. I really felt excited about the performance of postgres if we resolve some of the problems with it today and I saw similar expression on everybody else' face after the talk. The points Jignesh made will surely help postgres to achieve their goal for the next few years, I believe. Boy! I love dtrace!!
I wanted to attend search.postgresql.org to see how good the new search system on postgres website is going to be and it looks really promising. I am sure it will be very helpful for all the hackers and more useful for the novice ones like me.
I was very excited about the last talk "Multi threaded query accelarator". The idea of multi threading at the application level giving high overall performance with pg backend was certainly exciting. Not only that, they have also deployed such solution in production made me walk into that room instead of going to the lightining talks which I are always more fun. The talk presented detailed design of their systems and some analysis on how multi-threaded application could improve the overall performance. I personally thought, their database design was not too complicated which allowed them to get performance even if they added some complexity at the application level. Also, I felt, multi threading at application level is going to give performance boost of limited range for OLTP and data-warehousing applications. This is in a way obvious and got confirmed from the presentation to me.
I could attend some of the lighting talks and happy that I didnt miss out on DBIx::cache and dtrace ofcourse!
EnterpriseDB hosted dinner & drinks at a local pub and I must thank them for the nice food/desert and drinks they arranged and yes, a free Tshirt.
Socialising is always good !