One common question I guess is how much should I see performance improved in case I increase memory say from 16GB to 32GB. The benefit indeed can be very application dependent - if you have working set of say 30GB with uniform data access raising memory from 16GB to 32GB can improve performance order of magnitude by converting very IO bound load to CPU bound, it is well possible to see limited gains - if your working set already fits in 16GB you may not see any significant gains upgrading memory to 32GB. Interesting enough similar can happen for very large working set - for example if your main queries do full table scan of 100GB table it does not matter if you have 16GB or 32GB the load is going to be way too much IO bound anyway.

Interesting enough because of MySQL scaling issues it is also possible to see performance to go down as you increase buffer pool size. Some threads which would be safely sleeping waiting on IO completion are now …

For those unfamiliar with SQL Server clustering let me give you a quick blurb.  Microsoft markets SQL Server as having active-active clustering.  When most people hear active/active or clustering in the database they generally immediately think of features like Oracle Rac.  SQL Servers implementation of Active-Acitve clustering ( as of 2005 anyways ) is really a HA setup with one instance active on the A side, and a seperate unrelated instance active on the B Side.  These servers hove different ports, install directories,  and share nothing at the DB level.  A san is used with its disk presented to each node in the cluster.  Microsoft Clustering Services (MCS) is used to manage the IP take over,  handle resource transitions, etc.   The setup is really active-passive, passive-active.  This ends up confusing and even mis-leading people who are both technical and in management positions.

Why …

From LinuxWorld this week in San Francisco, me, Tux and Larry Augustin...

And then some time later...

I think someone spiked my drink!

Matt Asay writes today about the prospects for open source vendors going public or, more likely, being acquired, and wonders whether open source vendors should “hold out for an IPO” or “capitulate” and be acquired.

The latter seems far more likely, especially in the current economic climate. We have written before about the open source vendors most likely to go public in the next couple of years.

Looking at the list of contenders again it is easy to imagine that they could all be snapped up before they make it public thanks to the fact that 1) open source vendors are very attractive investments 2) it is difficult for open source vendors to build the momentum to do so.

I spoke recently with …

I recently spent 3 hours pulling my hair out trying to fix a MySQL stored procedure that wasn’t working as expected. I’ve boiled the ultimate problem down to a simple example below. Can you spot the problem?

# First, let's create some sample data
CREATE TABLE TrendicsTest (
DateTime DATETIME NOT NULL,
Value INT NOT NULL,
UNIQUE KEY DateTime (DateTime)
);
INSERT INTO TrendicsTest VALUES ('2008-01-01 01:00', 1);
INSERT INTO TrendicsTest VALUES ('2008-01-01 02:00', 2);

# Next, lets' define a stored procedure to query the sample data
DELIMITER |
DROP PROCEDURE IF EXISTS summarizeTrendicsTest|
CREATE PROCEDURE summarizeTrendicsTest(dateTime DATETIME)
BEGIN
SELECT @endDateTime := DATE_ADD(dateTime, INTERVAL 1 HOUR);
SELECT * FROM TrendicsTest WHERE DateTime=@endDateTime;
END;
|
DELIMITER ;

…

First of all let me start off saying that I learned a lot of Capacity Planning from two people. Jozo Dujmovic, and John Allspaw-who by the way is coming out with a book.

Capacity != Performance. You may have the capacity to do a bubble sort but a bubble sort is still a bubble sort.

Really to Scale you need to know when your application will break. I have a tool set to help determine what application is producing what SQL and use that to figure out which SQL is producing the most load on the system. Some common tricks I do is put the execution path automatically as a SQL comment, then sample the FULL Processlist to build a graph on what application, function, SQL pattern is the top load.

On top of that I use Ganglia to trend the use of each mysql …

As last year, MySQL will host three Customer Conferences in Europe. They will take place at the following dates and locations:

• Thursday, October 23, 2008: London, UK
• Tuesday, October 21, 2008: Munich, Germany
• Wednesday, November 19, 2008: Paris, France

The content differs slightly per location, but there will be sessions in two parallel tracks on a wide range of topics, including success stories from customers as well as talks on very technical/practical topics. Here are some excerpts from the agenda for the UK Event:

• Delivering Web 2.0 Applications with MySQL and …

The german print magazine Technical Review features two MySQL articles in his latest issue 09/2008. Kaj gets interviewed on the question what the future plans of Sun are with regards to MySQL. And I give answers to five technical questions on MySQL and upcoming features.

(...)
Read the rest of In the (german) Press: Linux Technical Review 09/2008 on MySQL (87 words)

The german print magazine Technical Review features two MySQL articles in his latest issue 09/2008. Kaj gets interviewed on the question what the future plans of Sun are with regards to MySQL. And I give answers to five technical questions on MySQL and upcoming features.

(...)
Read the rest of In the (german) Press: Linux Technical Review 09/2008 on MySQL (87 words)