Welcome to the 119th edition of Log Buffer, the weekly review of database blogs.
We start in the MySQL world with some engine news. On Brian “Krow” Aker’s Idle Thoughts, Brian explains the state of engines in Drizzle, the pared-down MySQL. He begins, “So many engines, and so little to choose from. This is one of our two major decision points in Drizzle right now.” Maria, Falcon, PBXT, and InnoDB are in the dock.
Arjen Lentz asks a simple question: Would you prefer InnoDB to be the default storage engine?, also the subject of a quick poll he created.
While we’re on the subject of defaults, Giuseppe Maxia, the Data Charmer reports on STATEMENT-based replication as the default in MySQL 5.1: “MySQL 5.1.29, the next (and last) release candidate, will revert the default replication mode to STATEMENT based . . . MIXED mode won’t be the default anymore.” The commenters are not totally happy with that.
Well, one of the good things about MySQL is that, if you’re not pleased with it as it comes, you can soup it up yourself. Giuseppe again looks at the community at work on SIGNAL — a feature scheduled for MySQL 6.1, but if you can’t wait for that to get SIGNAL and RESIGNAL, Giuseppe writes, you can build MySQL yourself with a patch by Jorge Bernal.
Likewise, Percona’s patches spread to a wider audience, thanks to the OurDelta project, ” . . . a community effort to provide builds with features (Percona patches, Google patches, etc) and storage engines (PBXT, Sphinx, etc) that aren’t in the main MySQL server.” This item comes from Baron of the MySQL Performance Blog.
The Google/Percona/OurDelta patches make another appearance on SmugBlog, where Don MacAskill’s gives a detailed report of his success with OpenSolaris + ZFS + MySQL in production.
The MySQL community also contributes tooling. Matt Reid announces the release of the Monolith Toolkit, ” . . . all of my various scripts together into a useful toolkit.”
Now here’s another question, this one from Mark Callaghan of High Availability MySQL: Who needs hash join — Intel SSD will soon be here. The nub of his query is this: “Intel SSD can do a random read in less than 0.1 milliseconds. It also has great random write performance which has been an issue for SSD and the price is reasonable. Decreasing random read latency by 50X to 100X will make disk-bound joins much faster.” The readers seem to think it’s not really an either-or issue.
The DB2 and Informix communities have some community stuff going on too. The 2008 International DB2 Users Group European Conference in Warsaw, Poland recently finished, and Robert Catterall covered it day-by-day, beginning with the first of his DB2 Notes from Warsaw.
On the DB2PORTAL Blog, Craig Mullins announces and sets the stage for his upcoming webinar on assuring the recoverability of your DB2 databases.
Moving the theme over to SQL Server, Jason Massie of StatisticsIO writes, “As much as I would like to be a SQL Villain, I got nominated as a SQL Hero. It was for my QPee Tools . . . ” — (“[a] set of tools to log SQL Server 2008 query plans, identify plan regressions and performance problems because of plan regressions.”).
On In Recovery…, Paul S. Randall announces his article on tracking changes in your enterprise database in TechNet Magazine. LessThanDot links to another Paul Randal article, this one on FILESTREAM Storage in SQL Server 2008.
Jamie Thompson, the SSIS Junkie, looks into a finer point of SSIS and embedded SQL. It’s a response to an item by Coastal Data Enterprises, Inc.’s Jon Jaques, SSIS and the Perils of Embedded SQL. Writes Jamie, it is a discussion of “[the] drawbacks of the Execute SQL Task â€“ namely that the Execute SQL Task has no understanding of the underlying schema[;] thus any changes will not get recognised by SSIS. [Jon is] quite right, the Execute SQL Task does not expose any metadata and hence SSIS does not validate the code within an Execute SQL Task . . . ” Jamie has raised a Connect submission, and he asks for you votes and comments.
My Oracle readers, you have been very patient. Here’s your Oracle stuff now.
The early-adopter release of Oracle SQL Developer Data Modeling appeared last Friday. Lewis Cunningham of got his hands on it and walks through his using OSDM to reverse-engineer a schema. His verdict: ” . . . I spent several hours with this today and never hit a single foobar. The program never died on me and I never got an error. I really put it through some paces too. I would also say, after having used more than a handful of different data modeling tools, this is a slick piece of software.”
Pete Finnigan announces an Oracle Security webinar: ” . . . on the 23rd of September at 15:00 UK time I am going to be doing another webinar on Oracle database security.” It’s billed as covering, “How databases are compromised (including live demonstrations) Techniques and tools that can help secure the DBMS Priorities in risk mitigation and remediation of security gaps”.
Bruce Schneier is kind of a big security guy, too. On his Schneier on Security blog, he discusses and links to an article on how to write injection-proof SQL. “It’s about time someone wrote this paper,” he writes. There’s lots of good comment here; example: “Once upon a time I used to think that DB Admins got paid so much money as compensation for the dullness of the job. These days I’m thinking it’s for the thankless nature of keeping upto date in their own time to keep the DB’s secure and up and running.”
Finally—Steve Karam, the Oracle Alchemist, advocates that we learn databases from Tico the Fairy!. Of course he’s talking about The Manga Guide to Databases. I predict an entire line of manga technical guides. The next one will be The Ghost in the Bourne Again Shell.
Okay, that is all. Please add your favourite blogs from this week to the comments, and send me an email to get Log Buffer going on your own blog. Till next time!