MySQL Performance Schema (7)
This is #7 in a series of blog postings about MySQL Performance Schema.
Seven is enough for now, I’ll try to wrap up this series of blog postings by replying to some of the questions and comments that I’ve seen recently.
Q: Will the same instrumentation be in all storage engines?
A: It’s too early to say. We have lots of instruments in Maria and MyISAM, a few in Falcon. A developer for another storage engine has shown interest. But technical issues exist and we’ll have to negotiate.
Q: Is instrumentation available for more than mutex and disk access events?
A: No. WL#2360 mentions other event categories but makes it clear that they are properly part of a subsequent worklog task. We are only guaranteeing that what is in WL#2360 will be available with first public release.
Q: When is first public release?
A: April 20 2009. This is the first day of the MySQL User Conference in Santa Clara (http://mysqluc.com). I regret I wasn’t able to secure a speaking slot, but I’ll be there, ready to attend a Birds of a Feather or talk in the corridors. Other MySQL folks have told me they’ll have things to say in their own talks, and I’ll blog about that when I’m certain.
Q: Shouldn’t we have heard earlier, after four years of work?
A: That’s a misunderstanding that’s probably due to a sentence in my earlier blog “I’m an architect so I did the original specification and most of the wrangling/wheedling/editing over the last four years to keep this target in our sights”. There were many objections to overcome, there was some customer discussion, and in between there were doldrums that could last for months. A speedup happened recently (I credit Sun Microsystems for this but the immediate cause
was that Marc Alff took an initiative). Sometimes this week I’ve felt that full disclosures like this blog series came too early, since of course they’ve stimulated people to express their impatience about waiting 10 more weeks.
Q: Is there something for memory, for locks, etc.?
A: Not now. What we do for WL#2360, and what we can most easily extend, is instrumentation of “events”. A look at memory structures or locks (including deadlocks) is instrumentation of “states”. We are able to show some state information, for example of open files, but sometimes state information is easier to acquire from a storage-engine memory, and for other things it will be easier to do more when we have upgraded our data dictionary, and that’s a separate project. Bottom line: not now, but we recognize that a Performance Schema is more than instrumentation.
Q: Is this open source?
A: Yes, it’s an open source cathedral. People who’ve read “The Tragedy Of The Bazaar” will doubtless remember that cathedrals were the largest community projects of a certain age, and will doubtless reflect that, when you’re trying to build something tall, you don’t succeed by stacking bazaar booths on top of each other.
Q: When will we hear more about this?
A: I’ll post in this blog again when events occur.
February 13th, 2009 at 11:03 pm
I think there are slots available to speak about this at the MySQL Camp or the Percona Performance Conference. A talk there would be great.
March 6th, 2009 at 12:46 am
Konstantin Osipov and I have reserved Thursday April 23 11:55am at MySQL Camp. Performance Schema will be only one of many new features that we demo.
March 6th, 2009 at 10:04 am
Hi Peter, quick question from a bystander: Google can’t find “The Tragedy of the Bazaar” for me… are you referring to Hardin’s “The Tragedy of the Commons”? Thanks!
March 6th, 2009 at 4:37 pm
Hi Ben,
Sorry, it was an insider joke. I’d just read a comment that this project smacked of “cathedrals”, which is a reference to an open-source manifesto “The Cathedral and the Bazaar”. So I whimsically decided to tout the community benefits of cathedral building, with a play on the title of Hardin’s famous article.
March 6th, 2009 at 6:51 pm
Ah, okay, yeah I’ve read “The Cathedral and the Bazaar” and so I thought you had a new counter-argument for it, which I hadn’t heard of. But it was a joke, okay.
Keep up the good work. PERFORMANCE_SCHEMA sounds exciting!