Saturday, 6 November 2010

PGWest: Day 3

PGWest; the final day. It started (for me at least) with Greg Smith giving another excellent talk on scaling PostgreSQL with Hot Standby. This was followed by Robert Hodges and Linas Virbalas talking about replication of data from MySQL to PostgreSQL using Continuent Tungsten.

The highlight/lowlight of the day - depending on your viewpoint - was Rob Wultsch's talk on MySQL. Rob primarily looked at some of the things MySQL does better than PostgreSQL, and also talked about the forks (or lack of them, if you discount patchsets - which I personally, do not) and the FUD. Now lets be honest here - Rob did make some perfectly valid points about MySQL; for example, the fact that it's replication is easy to setup. Now to take this example, I would argue that PostgreSQL isn't that hard to get going either - Robert Haas' tutorial illustrates that nicely - but MySQL is arguably better at the moment. For most of the points he raised, there are easy counter-arguments that can be made by PostgreSQL, as shown by JD who made a number of us cringe a little with an impromptu rebuttal session afterwards.

In my view, this whole session was doomed to failure. It's fine to point out some of the things that PostgreSQL can learn from MySQL, but the session as given glossed over everything that PostgreSQL can do but MySQL cannot - which, to someone unfamiliar with PostgreSQL could give an incorrect impression. For the end user who is selecting a database, it is important not to choose a product based on whether some features are better implemented in one database or another, but to choose based on the quality of the products, the reliability of them, and the availability of the features you actually need.

On reflection, I think the only way we can tackle this sort of comparison fairly in a talk session, is to have a proper, moderated debate between a PostgreSQL expert, and a MySQL expect. Who knows, maybe we can do that for PG East or PG Europe 2011.

Regardless of that - kudos to Rob for having the bottle to stand up and talk about MySQL in front of a room of PostgreSQL users. It was never going to be an easy crowd.

After lunch, I did a little work for a while, and then toddled off to Rob Treat's presentation on PostgreSQL 9.0: The other stuff. We joked the night before that no-one knew what The stuff was, never mind The other stuff, but I guess once he knuckled down to his slides, he realised it meant Hot Standby and Streaming Replication. The talk was pretty good in the end, though it did remind me a little of my talk in Brussels earlier this year, which could have been entitled PostgreSQL 9.0: The stuff, and the other stuff!

And that's where the conference ended for me, as we took the opportunity to hold an EnterpriseDB meeting whilst a we were in the same city. All in all, an excellent show, with a great turnout - and to top it off, The Register took notice of us with two different stories - which somehow makes it feel all worthwhile.

1 comment:

  1. Binary replication in PG 9.0 is awesome. It will be beyond awesome if they get sync replication with failover automation in the next release.

    I prefer fewer debates. I am not sure they will focus on things that are important to most users nor will they educate core team members about what the other side can do. My guess is that they will end up with silly benchmark results that matter to a one or two users. Someone recently published a result with 750k QPS on MySQL. That is an awesome result, but how many people need that?

    I hope there is a tutorial on PG for intermediate and advanced MySQL users at the next OReilly MySQL (and more) conference. I would attend that.

    I think it is best to let PG experts give talks on PG and MySQL experts give talks on MySQL and try not to mix the two.