Monthly Archives: July 2010

Parallel with 11G

Posted on by
1

Sigh….IF I could only live in “Uncle Larry’s” perfect database world where I have more control over what choices are made from the conception of the database’s life and know exactly what goals a development group has vs. Oracle.
How much easier a DBA’s life would be if we could just say, “Sorry, Oracle’s new feature would work well only if you completely redesign the logic of your process in code A, B and C of our main systems that have naturally grown in complexity over the years.  Of course, you don’t mind adjusting all priorities for the next six months while we accomplish this, right Management?”

The lead Mart Developer and I have been doing a bit of testing in a new development arena with 11g and in memory parallel tuning.  We created the database, the developer brought over the data he wanted to work with, starting with a smaller, 10% dataset to start with.  We set up the parameters just how Oracle’s recommendations stated, except for queuing-  We’ve tested this in another 11g environment and decided this was a poor choice until we are sure we have all hints in our old code out of the way that could impact with this parameter.

The mart developer removed all hard-coded hints from the CTAS statement, ensure he had only 10% of the standard data we usually worked with and executed the statement-  32 threads, excellent performance and we were thrilled! 

So thrilled, the developer pulled over all the tables the CTAS uses instead of the 10% we were working with.  Now, the thing that’s difficult to remember when going from hard-coded parallel hints to automatic parallel, is the development server has less resources than our production environment and more databases to compete with for resources.  Oracle is a lot smarter than we are and knows a poor choice when it sees one.  When it has the object and system statistics to know that performing a CTAS in parallel of this much data, with this little resources available, Oracle will make the right choice and no parallel is seen…

My lead mart developer was, of course, not happy and quite frustrated.  He wants to know why there isn’t parallel-  I review the explain plan and note in the statement that the process has been forced to serial.  Oracle has made it’s choice and it’s choice does not involve parallel.

Where does this leave a database administrator that has in the past explained to many that we were utilizing too many resources towards parallel that with 11g will have the new challenge of the database being a bit more conservative with parallel and being left to explain why you are not over-allocating, but not allocating at all?

At least I have a number of months to work with the developers to force more processes to run with an even lighter footprint on the database than has been done in the past.. :)

DBA Crushes

Posted on by
6

A little off technical topic and (hopefully) in good humor, I want to talk about DBA crushes.  Most techies have them, but as a female DBA, I can refer to them as “crushes” without anyone taking too much offense to the term, (hey, I’m after these guys for their technically brilliant minds, mind you!)   I’m sure some folks may bristle with the term, but we do have a tendency to follow certain “super stars” in the technical world like most people follow rock stars in the real world!

I’ve read DBA and Developer, consistently gush over Tom Kyte and Jonathan Lewis- two of the “alpha” DBA’s in the world of  database administration.  Many of the DBA’s, when discussing Tom or Jonathan, sound like my 12 year old daughter’s friends talking about the coolest boy in the class!  I do understand completely-  both Tom and Jonathan simply rock.  Now maybe it’s because I also didn’t go for the cool boys in school, but they don’t, no matter how technically solid and correct that they are, make me technically swoon…  I find their writing interesting, accurate and up to date, but they don’t make my mind race or give me the technical thrill ones I refer to as my DBA crushes do.

Now, my DBA crushes are the ones that I read an article or white paper from and it resonates with me.  There is a connection, a technical kindred soul that reaches out and filling the void of that technical feature, makes everything clear.  The light goes off in my head and I think “Finally, someone who understands!”  They are sometimes interested in the same aspects of the feature or simply explain the feature with  explanations and logic that synchronize with how my mind works.

Of course, I then proceed to technically stalk the person and search down every blog post, article, white paper and comment I can find from them.  The demand to reconnect and gleam more knowledge is always a powerful thing.   I am quite aware how I religiously read anything I can from Tanel Poder and Wolfgang Breitling on the cost based optimizer.  I search out Cary Milsap because common sense is just plain rare in database administration, let alone legibly in English.  For James Koopman and Tim Gorman I read their articles because I’ve worked with them, trust them- knowing that these are two great DBA’s that want you to be as great as they are.  I scope out Alex Gorbachev and Martin Widlake for all the various topics that make me go “hmmmmm…”, (which I like since my brain prefers to be worked out…)  Last, but not least, I really enjoy the no-nonsense view on Oracle by Niall Litchfield. 

In my decade or so as a DBA, I don’t know if I’ve met another one who didn’t have at least one DBA crush, one other DBA that they idolized and followed, either buying their books, reading their blogs or searching for their articles.  It’s a wonderful thing, these crushes, since DBA’s in the field are able to offer and connect to us in ways no one else can.   Honestly, my husband gets rather bored after hearing 20 minutes on the cost based optimizer and dynamic sampling… :)