PostgreSQL Conference 2008 and Party
Friday saw Japan's annual PostgreSQL conference held at the Izumi Garden Conference Centre in Roppongi Itchome. Billed as "Learn everything there is to know about PostgreSQL in one day" - the conference is the main event in the local PostgreSQL community.
With major organisational backing and sponsorship from SRA OSS Inc., Sun Microsystems, EnterpriseDB, NEC Software and others, the event had a distinctly corporate flavour to it. Along with the plush Roppongi location, I suppose this matches the image of PostgreSQL as the Oracle of Open Source Databases.
I have recently started as CTO of a company that uses an OSS stack, and PostgreSQL as the DB platform. With my DB background being MSSQL, Oracle and Sybase, I have some catch up to do, so this was a great opportunity to get some concentrated exposure.
Morning saw the keynotes, the afternoon was sessions ....
First Keynote was Bruce Momjian from EnterpriseDB. Bruce is a long-time player in PostgreSQL, and previously worked for SRA. He talked about the past, present and future of PostgreSQL, and how it was now close to parity with the major commercial vendors.
Next up was Ito-san from Sun Microsystems. I was hoping to hear a little about the MySQL vs. PostgreSQL situation, given that Sun is a contributor to PostgreSQL, but has just purchased MySQL lock, stock, and barrel. Unfortunately it was more on Sun's business model, which seems to be to encourage any software - especially open source - to use multiprocessor systems, which of course Sun will most happily sell you :)
I started off in the Community track, but that the session there was related to developing PostgreSQL - rather that developing on PostgreSQL. None-the-less, it was interesting to see how the community worked, and to realise how few people are actually at the core of development. Compare this to the teams that drive MS SQL Server or Oracle, and it is truly enlightening to see that focus and collaboration can do as much with a little, as money and marketing can do with much.
The second session was Fukushima-san from Seino Information Systems - the IT arm of transport firm Seino. He outlined how they use PostgreSQL as the basis for their system tracking GPS location signals that every vehicle in their fleet transmits every three seconds. They use this information of course to optimise their vehicle network, but also for governance, such as ensuring drivers do not exceed more that 4 hours at the wheel at a stretch, don't leave the engine idling, etc. The main topic of his session was how they are using Heartbeat + DRBD to provide redundancy for their DB system, and he gave a live demo of failover, by pulling the LAN cable out of a server. We watched as the sample app immediately began to draw data from the second server.
Next was another good session, this time with Watabe-san from NTT Data, who reflected on the trials and tribulations of providing commercial systems based on PostgreSQL. In the end, he said, the number of problems over the many projects undertaken over that period that were directly attributable to PostgreSQL was "Zero" - very reassuring to a novice user like myself. There was also some good advice from the field about PostgreSQL's unique VACUUM system for compacting databases, and how and when to apply it.
A contractor working closely with NTT, Yasaku-san gave the speech that NTT - or most Japanese firms - probably couldn't. His very entertaining talk about all the experiments he'd tried - and generally failed - to get PostgreSQL to do various things such as table partitioning, and in-memory databases. Another nice failure story was about replacing a disk in a very old RAID array. Apparently the disk and controller had - unbeknownst to NTT Data - used a firmware upgrade that was supplied by the disk vendor, who was no longer around. They were able to source the exact model disk from the manufacturer, but the moment they plugged it in, the whole system stopped - quite the opposite effect from what you'd hope from redundant technology. (Note to self - check RAID config for dodgy firmware before replacing hard disk!
The final session was on PGCluster, a truly Japanese open source initiative for clustering PostgreSQL, which - unfortunately - doesn't work very well. It was an entertaining story none-the-less, with lead (sole?) developer Mitani-san, telling the tale of how he stared the project, continued it after moving to the Netherlands, and of the different feedback and cooperation he had from various PostgreSQl communities around the world.
It was a highly educational day for me. There were a couple of criticisms - I'm not sure if this is supposed to relate to the Corporate image they were trying to present, but they really they could have used a cheaper location, with a little more space. Many of the sessions were packed out, the air quality stuffy, and the ceiling so low that only those sitting at the front of the rooms could see the lower half of the projection screens.
Still, after the event a good number of us moved to Roppongi for the excellent after party, and free beer, lightning talks and jan-ken-pon giveaways washed such thoughts away :)
