Doug's Oracle Blog

On the one hand, I regret the fact I seem to be posting a lot of links to other people's work at the moment but I couldn't let this one by, if only for those who don't use OraNA or the like and might miss it.

The one consistent message I kept hearing from other attendees at Openworld this year was that a lot of sites seem to have run into performance problems when upgrading from 9i to 10g due to changes in the CBO and DBMS_STATS. Yes, there are some Oracle 9i databases out there, you know.

I noticed via the Optimizer Development Group's blog that there's a white paper on this topic (PDF) on OTN by Maria Colgan which summarises some of the issues in a very readable form. (Well, it could maybe have used one more round of proof-reading for missing words etc., but I'm being picky.) It's not particularly deep but I'm glad to see it, if only to highlight some of the issues that you might face and can then investigate in more depth on your own systems if you think they'll affect you. Even if you know the issues already, it might be useful revision when you're about to upgrade to 10g. I know that I constantly come across people forgetting the automatic GATHER_STATS_JOB when they work with a mixture of versions and as Noons pointed out in the comments on this post, that can really throw things if you're not careful.

(Of course, there's a funny side to this too, in that it discusses the new default approach to Histogram generation that caught me out the other day. Well, it wasn't so much the new approach as me being hasty and seeing what I wanted to see, but the article does discuss the difference in Histogram generation using the default METHOD_OPT!)

It also mentions the change in the default value of parallel_max_servers that I've seen more than one system suffer from and discusses a SQL Test Case Builder that will be in 10.2.0.4 to help in submitting Metalink SRs. I'll have to play around with that at work because there are times when building a test case has proved a pain in the behind (as I'm sure I remember Peter Scott blogging about in the past).

Here are a few interesting blog postings I've been reading this week.

Jonathan Lewis on another Statspack example.

Nuno Souto on SQL Server - in three parts. What I particularly like about these postings is that the man has used both and is honest about their relative strengths and weaknesses, rather than sticking to the party line.

Similarly, Raj Thukral on MySQL and, despite his Oracle background I know that Raj is pretty positive about MySQL in general. (I almost called him a MySQL fan!)

Alex Gorbachev admires "Kurt, the Dude guy's" tool (so to speak).

Kevin Closson has so many interesting posts at the moment that I'll just point you to his blog again

(Now, which blog does this posting belong in, again? I've forgotten already ...)

Some of you will already know this, but ....

I pointed out the following entertaining thread on one of the OTN Forums to a friend. He asked how I had noticed it.

I'm sorry to disappoint you, but I don't spend all day on the forums (and at this point I should introduce my recurring promise to blog about that one day). I just happened to notice that Jonathan Lewis had updated a couple of his articles. Despite the fact that a lot of his activity is focussed on his blog these days, don't miss out on updates to the articles via this handy feed.