Friday, September 29, 2006

Log Buffer #12: a Carnival of the Vanities for DBAs

I fell for it. I commented on the Log Buffer and shortly after that I was offered to host an issue. Here we go, then. Welcome to Log Buffer #12!

Confessions of an IT Hitman by Dratz features a strong message: Don't build a data warehouse, arguing that most of the times a DW is built, it's just because a clueless customers was either listening to buzzwords from a salesman or following the latest trend. The key message, for the ones who missed the build-up of the data warehousing movement during the past decade, is DW is a business solution, not a technology solution. OTOH, there are projects that would really need a data warehouse, and don't get one. But this is a different story.

Mats Kindahl in his MySQL Musings talks about Replication and the disappearing statements, i.e. the risks and gotchas of replicating data in MySQL while limiting the databases involved in the replication process. It is something that the manual states quite clearly, but it bites back quite often nonetheless. Mats explains some unusual points in this old issue. Querying a table with a fully qualified name may result in more than overkill. You may be the next victim of the infamous disappearing statement.

Brian Aker is always a volcano of ideas. His Brian 'Krow' Aker's Idle Thoughts don't show much idleness. Rather, he's often producing some hack at great speed, or he's pondering on a new hack. This one, Firefox, yum, sounds interesting, because the suggestion came from MySQL CEO, who is also a man of vision. It looks like something will come out of this clash between dreamers.

Vadim Tkachenko from MySQL Performance Blog, which he runs together with Peter Zaitsev goes int a wrestiling with the latest beta of SolidDB for MySQL.

Test Drive of Solid is a comparision between the behavior of the SOlid engine compared to InnoDB, to check if a migration from InnoDB to Solid could be done easily. Vadim finds several differences and a few bugs. SolidDB doesn't seem to be solid enough in this beta, but it's getting closer.

A charming set of slides from Lars Thalmann in his Lars Thalmann's Blog talking about the joy and the shock of Being a MySQL Developes. If you thought that a developer for an open source company had an easier job, think again. Go through Lars's slides and discover a new dimension of cooperative work.

Good news from Markus Popp ( blog). For almost one year I have waited for MySQL status and variables to be available as INFORMATION_SCHEMA tables. Now it seems that MySQL 5.1.12 offers New information_schema views that do exactly that. I built MySQL from source a few days ago, and the new features were not there. After seeing Markus's post, I rebuilt it, and spent some time toying with the fresh additions. For example, calculating the key cache efficiency is one task that previously you had to perform with the help of a programming language. Now you can do it in SQL.

A glimpse into the future from Mikael Ronstrom, senior software architect at MySQL AB. The State of MySQL partitioning seems to be close to bug-free, and he's already fiddling with new enhancements in the next version of MySQL. It's almost a pity that I had to submit a new bug for partitioning just hours after his post!

Matt Asay from Open Sources examines The spirit of winning, taking into account the elements tha make a winning team in athletics, and finds that the same elements apply to open source competitiveness. Food for thought.

Bob's World by Bob Field is usually full of interesting concepts. This one, Advanced Features as Crutches is no exception. Advanced features, Bob warns against using an advanced feature just because it's there. Everything has its place, but it is no mandatory to use all the frills a DBMS engine offers. If you do so, security and simplicity may suffer. I fully agree.

In The Oracle Base Blog, Timothy S. Hall turns a joke into a lesson. So many people ask Where's the 'Go faster' switch, and they won't understand the hard work necessary to actually improve the performance. So he makes a surprising proposal. Read on.

Craig Mullings's features two entries that appeal to common sense. Intelligent Database Management lists the tasks of a DBA, stressing the effort that has to be put into the job. Only through intelligent automation of the tasks it is possible to keep up with the chore of managing database systems at a high professional level. Choosing the best DBMS is an old question that does not get a straight answer. Craig goes explains what you need to take into account to reach a sensible decision.

If you liked the previous entries, be aware that Craig writes in different places. DB2PORTAL Blog is one, and there he wrote Sequence Objects and Identity Columns. It is not another chapter of the natural vs surrogate keys saga, but a lucid explanation of what are the differences between identity columns and sequence objects, when each of them is needed and how to handle them. Even if DB2 is not your database of choice, his explanation can help you choose the right solution.

This entry gives a deja vu feeling after you read the one by Craig about Intelligent Database Management. One day in the life of a DBA, as seen by Jeff Hunter is what you get from reading What do you do all day? (a piece of So What Co-Operative). Not a common DBA, mind you, but a very organized one. As seen previously, the good DBA job is a balanced mixture of experience, organization, and an educated choice of technology. That way, even long lasting tasks can be approached with a quiet frame of mind. Jeff's tasks include such things as monitoring the databases, insuring that backups are done, checking for critical conditions. He sounds less worried than the average DBA, because he merges experience and knowledge with the right tools for the job.

Raven Zachary is a contributor to The 451 Group. In Open source database poll highlights barriers to adoption he comments on a survey launched by SolidDB about the acceptance of open source database. It seems that most of the IT professionals attending a recent Linux Expo were timid about adopting open source database.

A different angle is offered by another research, as reported by Zack Urlocker in his The Open Force blog. Linux Database Attach Rate says that 85% of RedHat customers show a strong interest in open source databases. I guess that this research and the survey in the previous item come from different sources. Zack also reports on the Zmanda Recovery Manager, an open source (dual licensed) tool dedicated to backup and recovery of MySQL databases.

Paddy Sreenivasan covers the same topic in Selective recovery of MySQL databases at the O'Reilly database weblogs. Paddy highlights the common needs in a backup solution and explains how Zmanda covers them.

If you jump to Sheeri Kritzer weblog, its title "My-ess-queue-ell vs. My-see-quel" will give away her main area of expertise. One intriguing entry in her blog, Choosing datatypes for fields is a clever hack to improve the quality of data uploads into MySQL tables. So simple, and yet so good. To choose the best datatype, first use a bad one, and then let the system tell you which ones you can have. Well done! If you know Sheeri (I had the pleasure of seeing her at work during the MySQL Users Conference, but reading her blog is enough to get the idea) you will see that she doesn't accept anything blindly, but she needs to explore and dig deeply every concept. In Isolation & Concurrency she takes Jim Starkey's assertion that MVCC could not be serialized as a starting point to an enthusiastic explanation of why you can't do it.

DBMS2 features a piece by Curt Monash, delving into the intricacies of data warehousing terms.
Data warehouse and mart uses - a tentative taxonomy is a quick introduction to the world of business intelligence by comparing the features that each different technology supports or requires. It won't replace a good book, but it sets the record straight in less than a page. Really commendable.

The scintillating Roland Bouman has done that again. He saw an intriguing problem with some clever solutions, and found a way of doing the same job in pure SQL. The matter at hand was Finding Redundant Indexes using the MySQL Information Schema. Others have done this before, using different methodologies. Roland points to the language-independent solution, and he delivers, as he usually does. Browsing Roland Bouman's blog you'll find several witty examples of his analytic approach.

A storm of advice for Oracle developers from ... a DBA. Andy Campbell in his Oracle Stuff I Should Have Known ! gives sensible advice to developers. If Only ... more developers used some nifty features that andy explains in full, their life (and the DBA's as well) would be much easier. This kind of advice (use application metadata to say which parts are being executed) would be good for any DBMS. If only ...

Thomas Kyte grabs the suggestion in The Tom Kyte's Blog and elaborates on his own about Action, Module, Program ID and V$SQL.... This entry, like the previous one, generated quite a stir, with good and bad vibrations on both sides of the Developers-DBA divide.

And finally, a personal closing note. I mentioned in one of my previous posts that I joined MySQL AB as QA developer. A few days in the job, and now I am experiencing the thrill of working in a virtual company

1 comment:

Mike Kruckenberg said...

Nicely done Giuseppe. Glad to see other MySQLers getting involved.