PostgreSQL Summit Day 2

Yesterday was the last day of the conference… I have to admit the time really did go by fast. I started the day in Tatsuo Ishii’s [http://pgpool.projects.postgresql.org/ PgPool] talk. He gave an overview of pgpool, how it works, and some ideas that are being worked on for pgpool2. This piece of software actually has quite a bit of functionality to it, a key piece of which is a simple replication solution that is akin to the replication system that mysql offers, only not built in but also a bit more robust. I’d like to see more people putting time into it on this side of the world, I think it really could be a good solution for a lot of people. After that I went to Andrew Dunstan’s [http://www.pgbuildfarm.org/ Buildfarm] talk. Most of this was pretty well known to me, but was interesting to get some of it from the horses mouth. There was also some talk about future features people would like to see, including some talk of updating/revamping the website. I’m a little twisted on that end… I would like to help push that piece forward, otoh the site really is for the developers, so I’m not sure it needs to be that wizbang; we might better focus on some other things. After the break came a series of 4 talks known as the NTT Power Hour. [http://www.ntt.co.jp/ NTT] is a major telecommunications firm in the far east, and, as it happens, they are doing a whole lot of stuff with PostgreSQL as the basis. Some of the things covered included a new checkpoint algorithm that basically used the concept of a slow writing checkpoint that would use all of checkpoint timeout time to write, eliminating the negative performance spike that normally occures at checkpoint. There was some confusion on how this differs from the current background writer work, but a patch should be submitted soon at which point all should become clear. Another presentation was a high speed data loader that was created to replace COPY. Apparently they are processing obscene amounts of data (like 800+ GB a day or some crazy number like that) so they needed something that could push data around faster, but it really had a lot of caveats, and a few people asked about how those might be worked around. I think those questions may have been a bit off the mark. My thinking is that those people were looking for a general replacement tool for COPY, but what I feel this was a better example of was how PostgreSQL’s flexibility can allow you to create custom tool relatively easy (this was just a custom C function) to achieve amazing results. The two other talks took that idea and kicked it up several notches; one was on an incremental backup system that bridged the gap between pg_dump and PITR; the other on a system to do parallel scaling with postgresql. Of the two I thought the incrementalbackup looked far more useful… the horizontal scaling didnt really see any advantages until you got above 6 servers, and the complexity involved didn’t seem worth it, however with refinement they can probably imporve on that, and if you have alrage enough system I can see it being used. On other note on the NTT talk… the presenters themselveshad really limited english skills, but I have to say they all did an excellent job with putting thier ideas across and keeping people interested… this to me was one of the real highlights of the whole conference. After the power hour, Devrim grabbed a few of us for lunch. I was a bit concerned as we headed down the back alleys of Toronto headed for some myseterious turkish restaraunt, but it turned out they had really excellent [http://en.wikipedia.org/wiki/Shwarma shwarma’s], which are very hard to find in Florida, so I was really pleased with that choice. After lunch I swang into Peter’s XML talk. He did a really nice job of defining what the actual problems are when people talk about XML support, and gave a number of pointers to some standards information that people should be looking at. We also discussed solutions a little bit, though things didn’t really get solidified, but hopefully once Peter’s information get’s out to the greater postgresql community we’ll see some movement on it. After that I ended up going to the tsearch2 developement talk. It didn’t really do much for me to be honest; it was basically a technical review of what tsearch is and how it works; all certainly was good information but stuff I was familiar with. What I really was hoping was that there might be some discussion of how we can make tsearch simpler to use for end users, how we might go about getting more languages support, more dictionaries into the mix, and make stemmers more easily available for other users. Barring that, I spent some time scrambling to find a world cup live tv feed, and luckily by the end of the talk I was in (many, many, thanks to “PrimerTV”). This did kind of interfere with the Web Team meeting, but it was small and mostly just a discussion amonst the regular contributors on some housekeeping we ought to be doing. I kept one eye on the futbol match most of the time… it was certainly the more exciting of the two. [[image /xzilla/templates/default/img/emoticons/wink.png alt=”;-)” style=”display: inline; vertical-align: bottom;” class=”emoticon” /]] After that was a parting presentation with words from Josh and Bruce on PostgreSQL. More importantly the match had gone into penalty kicks, of which the one miss by the french was amazing to see. I noticed we didnt actually have any italians at the conference, but congrats to any who made the choice to stay home for the team [[image /xzilla/templates/default/img/emoticons/smile.png alt=”:-)” style=”display: inline; vertical-align: bottom;” class=”emoticon” /]] After the conference closed we had a large group photo outside, and a quick exchange of some gpg keys. It’s too bad that wasn’t organized, it could have been a great thing to spread the web of postgresql trust, but maybe next time. After that a large group of us went to dinner where there were a lot of interesting conversations to be had. One thing that we all seemed to agree upon was that, as posibly the only native english speaker who also kows japanese and postgresql, someone should really hire Michael Glaesemann to spend dedicated time bringing us information and code out of the Japanese Community. After some time there I swung back to the hotel and got distracted by some code before turning in for the night. I’ve decided to blow of the [http://pgsprint.wetpaint.com/ code sprint] in favor of some other things before I head out today… the weather is a bit miserable which is probably a nice way to remind me of why I live in Florida. (At least until someone get’s me that job in Australia [[image /xzilla/templates/default/img/emoticons/wink.png alt=”;-)” style=”display: inline; vertical-align: bottom;” class=”emoticon” /]])