Jun 14, 2012 / By Marco Tusa
These past couple of weeks have been quite intense and generated a lot of news.
What stood out to me was the escalation in the race between MySQL/Oracle and MariaDB to be better. I have to say that all of us are benefiting from this competition. I am also not fully sure that Oracle would have put in such effort if Monty and his team was not pushing so hard. Certainly, Thomas Ulin has his agenda, but Monty is doing a great job for his product and in pushing the competitor to do his best. The result is that MySQL is becoming a better product day by day.
Is it a crazy idea to use MySQL for Json and HTTP embedded? I love crazy things, and this seems crazy, but it is closer to our reality than we can imagine:
Group commit is an important step ahead, not only for performance, but also for data consistency. I was really exited to see all the improvement from both MySQL and MariaDB. This is very good news.
Now, consider moving from accessing TABLE_SHARE to multiple TABLE cache objects, reducing the usage of LOCK_open, and implementing performance. This change is, again, part of the optimization process on the way MySQL internals works. It might seem minor, but it is not.
MySQL is locking ALL partitions in a table before applying the partition pruning. This has the effect of slowing down a lot the performance when a table has a large number of partitions. Obviously, this is not the expected behavior, and the fix will be in MySQL 5.6.
I am really wondering how many improvement were made in the 5.6 version. Anyhow, as a colleague of mine said: “Finally!!!”
Related bug: Partition bug fixed http://bugs.mysql.com/bug.php?id=37252
Hurray!!!!!! This is the kind of news that makes my day much better: ”Foreign Keys in MySQL Cluster”
I was jumping on my chair when I read it. Many customer had complained because they were unable to use FK and then chose something other than MySQL Cluster.
I will also soon test the performance.
MySQL Managing data
Monitoring performance and getting statistical data, I was always concerned about the cost of enabling it on a production system. I am glad to finally see some information on this topic. Receiving additional information on what is going on on my MySQL server is always good, and honestly the way we currently get info, or rather the way we don’t, in MYSQL for users is frustrating.
Tungsten & Vertica
The approach described by Robert Hogges is not bound to Vertica only but to any data storage that can accept the batch load from CSV format. Some interesting points are the data transformation, the possible parallelism by SHARD, and the error handling. Last but not least, there is the data transformation that can take place.
On the subject of MariaDB & MySQL, I found the calculations/test done by Axel interesting. Of course you need to use SSD, but some of the findings could be use to have a more effective setting in MariaDB.
And then something made me laugh. I need to test and confirm the numbers, but if it’s true, it would be really funny.
What is so funny, you ask? Well, Oracle is trying to place MySQL as the MS SQL Server competitor, and suddenly MariaDB comes out as a better product on Windows. Will Oracle try to buy MariaDB? Should we just need to wait for MySQL 5.6?
I’m thinking of going with the second one. This race to always be a step ahead of the other is undeniably interesting. MySQL is becoming a more solid and trustful product because of it, which can only benefit the customers.
A good read:
Close to MySQL
Will trying Cloud storage engine be worth it? Working with Amazon and having more and more customers asking for it, I found that this solution was worth a test. I plan to do so in the following month.
In the mean time, I suggest the following reads:
Leave a Reply