For some reason my posts aren’t getting over to the planet site, so here’s a test post before I actually write something that might be important.

For some reason my posts aren’t getting over to the planet site, so here’s a test post before I actually write something that might be important.

We're making steady progress in removing bottlenecks in the Drizzle code base. So far, a number of mutexes have been removed and we've begin to replace a number of contention points with atomic instructions which remove the need for a lock structure on platforms which support atomic fetch and store instructions.

I'm pretty positive about the direction we are going so far. We're seeing the right trends in our scaling graphs, with very little performance drop off in read-only workloads up to 4X the number of cores on the machine, and little performance drop off on the read-write workloads up to 2X the number of cores, as you can see from the graphs below.

It's a little difficult to see, but we've made a small but steady improvement from r950 to r968, with numbers increasing around 1-2% across most concurrency levels. You can see the raw numbers here:

+--------------------------------+-------+-----+---------+----------+
| …

Looking for a summer of code project for a database?

http://drizzle.org/wiki/Soc

We are looking at projects for the server that we are actively interested in having in Drizzle. Expect that your project, if successful, will be included in the current release of Drizzle.

I don’t know if it is because of my science background, I am a physicist, I do like graphs, especially when I do performance tuning. With UNIX like operating systems, the vmstat command give you an easy way to grab many essential performance counters but, generating graphs from vmstat output with tools like OpenOffice Calc is time consuming and not very efficient. In order to solve this, I wrote a few scripts using gnuplot but they are not very easy to work with. Then, doing some benchmarks with DBT2, I found the vmplot.sh script and… I like that one. I just hacked it little bit to make it keeps the graph on screen, adding the “-persist” parameters to the gnuplot invocations. The script will produce 7 graphs that will be displayed on screen and save in png format in /tmp. The graphs it produces are the following:

• CPU: graphs idle, user, sys and wait time
• …

So we have completed a major new version of libmemcached:

Added new UDP fire-forget mode.

This is in line with the model that Facebook uses. We have added a fire and forget mode for operations which can just be "sent" to the server. Depending on your client application and the size of your data you may be greatly surprised as to how much performance you can get out of this.

Reworked performance for mget() to better make use of async protocol

We have known for a while that our mget() operation while being fast, could be made faster. Instead of now cycling through servers we poll on data coming back from the server. This allows us to skip more "wait" cycles while data is being sent.

Cleaned up execution of fetch (just one set of code now)

This moves us to a cleaner code base. Some of the same thinking we have been doing in Drizzle. Create …

This Thursday (April 2nd, 14:00 UTC), Patrick Crews will give a MySQL University session on How to Create a Test Case. This is an updated session of a talk we had in 2007, but this time it will be recorded (slides and audio). Patrick is a database engineer who works in the server QA department, so test cases are his daily bread.

For MySQL University sessions, point your browser to this page. You need a browser with a working Flash plugin. You may register for a Dimdim account, but you don't have to. (Dimdim is the conferencing system we're using for MySQL University sessions. It provides integrated voice streaming, chat, whiteboard, session recording, and more.) All MySQL …