Log Buffer #113: A Carnival of the Vanities for DBAs
Sep 5, 2008 / By David Edwards
This is the 113th edition of Log Buffer, the weekly review of database blogs.
Sheeri Cabral gets things going this week with her coverage of this rumour: Monty Widenius Leaves Sun/MySQL. If it turns out to be true, that seems like bad–or at least sad–news for MySQL, but Sheeri’s take is mostly an optimistic one.
MySQL’s Kaj Arno responds to the rumours on Monty resigning: “First, it’s a rumour. . . . Second, Monty’s resignation has been a possible outcome already since years before the Sun acquisition. . . . I can neither confirm nor deny the rumour.”
On CNET’s The Open Road, Matt Asay says. “Monty has done the right thing with his dissent. He has taken it outside the company, as Arjen Lentz, MySQL’s twenty-fifth employee, did before him. Arjen continues to be both a promoter and critic of MySQL, but is able to do so publicly without the constraints of an employee agreement. I assume Monty will do the same, and rightly so.” That’s not all the response out there, and there will certainly be still more as this develops.
In non-Monty blogs, Brook Johnson of Database Science asks, can a timestamp be slower than a datetime? He begins, “One should use timestamps rather than datetimes as timestamps are four bytes and datetimes are eight bytes, right? . . . When I was trying to quantify the improvement . . . I found the exact opposite – when used as a criteria, a timestamp is slower than a datetime. Often much slower.” As his tests demonstrate.
Sergey Petrunia makes his EXPLAIN CONDITIONS patch available. He writes, “I’ve made a patch that makes EXPLAIN show conditions that are attached to various points of the query plan.”
Explain that. Analyze this! Peter Zaitsev of the MySQL Performance Blog also has a go at some common wisdom. He says, beware of running ANALYZE in production: “ANALYZE TABLE should be very fast and non intrusive operation doing just little update on the data. Right? Wrong! There is the bug or rather MySQL Design Feature which causes ANALYZE TABLE to block all accesses to this table while it could be flushed from the table cache.”
On Opinions, Expertise, Passion, Ronald Bradford tackles naming standards. “If you were going to create a MySQL Naming Standard you have to make a number of key decisions. Generally there is no true right or wrong, however my goals tend towards readability and simplicity. In 2 decades of database design I’ve actually changed my preference between some of these points.” He shows how he now does it. This is the kind of thing we geeks just lo-o-o-ve talking about, so naturally, Ronald’s readers have a lot to say on this too.
Keith Murphy has issued his call for articles for the Fall issue of MySQL Magazine. Likewise in The Aquarium, Giuseppe Maxia emits the call for papers for the MySQL Users Conference and Expo 2009.
To Oracle blogs now. Tanel Poder gets it going with the second part of his series on Oracle hidden costs revealed: using DTrace to find why writes in SYSTEM tablespace are slower than in others. A long title, and a long, in-depth article. Worth the reading.
Of course there was some browser news this week, too. No, not that news. This is provided by the Ubiquity Firefox extension and by Eddie Awad. The item is Eddie’s look at new ubiquitous ways to search Oracle documentation and more.
In SQL Server land, MS’s Dan Jones has a worthwhile item, PowerShell vs. T-SQL or Why Did We Add PowerShell Support in SQL2K8.
Sticking with MSDN for the moment, PSS SQL Server Engineers offers How to Rebuild System Databases in SQL Server 2008. It’s worth a read, for as the item explains, compared to the process in 2005, “ . . . the command line switches have changed some and the process behind the scenes to rebuild the system databases . . . is also a bit different. Currently the SQL Server 2008 Books Online only mention an option for setup called /REBUILDDATABSES . . . but this information is not correct so I’ll outline how to do this here in this blog post.”
Kalen Delany had the BOLs to ask, Did You Know: Specifying Checkpoint Duration. “I realize probably no one reads Books Online cover to cover, so there are probably lots of little details that easily slip by you. I was introduced to one of the hidden ‘features’ just last week. Here it is: CHECKPOINT can take a parameter! . . . Maybe I’m missing something,” she continues, “but this seems like one of those features that was added ‘just because they could’.” And her readers shed some light on this shady corner.
Kevin Kline has some tips and links for Counting Context Switches, PerfMon Counters, And Other Miscellaneous Notes.
I can see by the clock on the wall that this is all for now. I’ve short-shrifted PostgreSQL this week, but I promise that Nicklas Westerlund will make up for it next week, if I have anything to say about it. Till then!