Drizzle7, a Rackspace backed MySQL fork - The icing on the cake | Hacker News Reader

Drizzle7, a Rackspace backed MySQL fork - The icing on the cake

linuxjedi.co.uk

82 pointsLinuxJedi31 comments

Remember there is also MariaDB, quote from they webpage:

MariaDB is a database server that offers drop-in replacement functionality for MySQL. MariaDB is built by some of the original authors of MySQL, with assistance from the broader community of Free and open source software developers. In addition to the core functionality of MySQL, MariaDB offers a rich set of feature enhancements including alternate storage engines, server optimizations, and patches.

http://mariadb.org/

I have a lot of respect for MariaDB, but they are aiming to be a drop-in replacement for MySQL. Drizzle is not a drop-in replacement and intentionally broke some compatibility to reach it's goals.

Not denying it, both projects have slightly different goals. But since some people reading this discussion might consider switching from MySQL to Drizzle it doesn't hurt to present them other options, which in some cases might suit them better (depending on what are your exact needs).

The second part is already online although not linked form the first one: http://www.linuxjedi.co.uk/?p=152

Edit: Actually it is linked ("Next post") but I was looking for it on the text itself, sorry! Also, my link was from a Google search, edited to the correct link.

Thanks for pointing that out :) The final part will be up tomorrow outlining the future plans for Drizzle.

One of the best features in drizzle for me is its flexible replication, i built rabbitreplication about a year ago which replicates from drizzle to basically anything over rabbitmq.

* note that rabbitreplication needs a bit of love to actually work since the GBP messages have changed since i last had time to update it.

We do plan to work on things like Rabbit and Tungsten replicators soon. In the mean time we recommend the native solution which is similar to MySQL's master/slave but uses the GBP messages stored in an InnoDB table.

What is the advantage of dropping the smaller integer types? Does it reduce the per-record overhead for example? I have worked with a table that had about 200 million rows and 4 1-byte tinyint values. If those had to be full 4-byte ints it would require another 2GB to store and reduce the number of records that could fit in memory. Unless there is savings elsewhere?

I think the point was to trim down the beast. For that you need tradeoff some features I guess. So small int were cut off. As were triggers, views and stored proc. Most of which is hardly used by target audience, webdev that is.

There are plans to bring back at least stored procs very soon. But to do them properly and as a plugin type (so if you don't need them they aren't there)

With InnoDB I believe that it stores the smaller int types as 4 byte anyway. But I am not an expert on this particular section code.

In theory you could pack the values in yourself into the larger columns.

It seems like the draft of NoSQL database solutions has made people think about databases, and thus also bumped up innovation in the SQL domain. Many interesting projects around MySQL these days.

Is there something similar to phpmyadmin that will work with drizzle? It seems phpmyadmin is not 100%. http://blog.cihar.com/archives/2010/10/21/drizzle-phpmyadmin...

The link you supplied is the current effort to port phpMyAdmin. I'm not aware of any other similar projects right now. It would be really good to have such a thing though.

Can someone point me to some good benchmarks showing the performance of Drizzle7 vs. MySQL 5.5?

Unfortunately we don't have any yet, although there are libdrizzle vs. libmysqlclient benchmarks floating around.

We don't benchmark against MySQL ourselves because we can very easily bias it in our favour. We would love to see other people benchmarking.

We do, however, benchmark every build against ourselves using several different suites as can be seen at:

https://launchpad.net/~drizzle-benchmark

Isn't there any standardized SQL benchmarks? I remember one open-source.

edit: http://sourceforge.net/projects/osdb/

We use SQLBench, Sysbench (I believe both are pretty standardized) and drizzleslap. I'll ask others in the team about the one you linked to, looks interesting.

Oh, I forgot we run DBT2 too

Has anyone tried this in a production environment ? Stable enough ? Any gotchas ?

We don't use Drizzle directly, but have been using lib-drizzle to give us c library access to MySQL. In almost a year, we haven't seen any negatives.

A good point, quite a few people are using the libdrizzle client library bundled with Drizzle. It is a high-performance BSD licensed library that can speak the MySQL protocol.

beyond blogs (my wordpress blog included), forums and mail servers I don't believe it is used in a production environment yet. But Rackspace intends to use it in production environments.

We believe it is stable enough (hence the GA) our testing and regression suites are very comprehensive.

Very interesting to find this article today. I have been looking at MySQL replacements and drop-ins recently, having looked at Percona and MariaDB. I'm definitely going to keep my eye on this.

Is this faster than postgresql?

If you defenestrate it, it accelerates at about 10 m/s^2.

This is also true for postgres, apples and feathers.

Faster is relative, Drizzle uses InnoDB as a storage engine and I suspect Postgres and Drizzle/InnoDB have optimizations for different workloads.

I would love to see benchmarks of things like OLTP workloads though.

Deleted comment

Drizzle7 is the GA release of Drizzle

Reply on news.ycombinator.com