A DBA friend of mine recently wrote me asking for help with the Unix “display” variable setting. We emailed back and forth for several days before he figured out that he had the wrong installation packages for his OS. After about two weeks of back and forth, with the right packages, I had identified the fix on the first day. He had not properly used the “xhost +” command in his setup.

We also talked a lot about Amazon and virtual machines, and how easy it is there… click, click, click, click… and you’ve got a machine.

My friend was eager to do the installation on some physical hardware he had, and is actually one of those skeptical people who doesn’t trust the cloud. He did understand my arguments for using Enterprise Edition, Enterprise Manager and the extra cost management & tuning packs.

After successfully installing the database, his next “endeavor” was to install EM. After reading the documentation, he said “Wow, there’s a lot of both pre-installation and post-installation steps that a person needs to do in order to accomplish it! The number (and difficulty?) of the steps appears to rival the task of installing a database itself! ”

He wanted experience with using EM… my most recent email to him follows:

As I said before, the last distribution of 11gR2 that I did included both EM and the management packs. It was all installed with the DB… I had never seen that before, and yes, it is a bear to install, configure and use. If you don’t have the the management & tuning packs, it’s not a very useful tool. But, it is the ONLY way to go. People who use Toad as their primary interface do not have a clue as to how to tune an Oracle DB. Quest products do not use the OWI… aside from the fact that you have to load trace files manually. They only require trace level 4 to load; but, OWI counts are not included unless you create them at level 8 or greater (12, or 16). Quest products will let you load any level you create; but, it doesn’t use the “wait” data. They still tune using the 35 year old techniques of ratios… hit ratios, percentages…. We know now that these ratios can be exactly at the recommended value and the database is not tuned, can be the worst performing DB in existence. Based on casual observations of Toad usage… and other quest tuning products… I put myself in the top 5% for tuning… both the database and the 3rd party application SQL. Setting the DB up properly to tune itself continuously, as it is capable of doing, does truly turn the “Oracle DBA” into a commodity. The DB and the code will not need us forever. Learning the Oracle interface and the tools available is currently the ONLY way to go.

When you deal with “waits”, the internal counters built into the kernel of the DB, you have the exact value of any time waited, and in theory, the reason it waited. You take the waits one at a time and eliminate them. From a performance perspective the DB is either running as fast as is physically possible, or it is waiting. Eliminate the waits and the DB is “tuned”… there is no slow, medium or fast. It’s running full speed, or it’s not running.

Tuning is a breeze when done properly. Enterprise Edition. EM and the required packs… that is the solution to performance.

Now, if you want to run in standard edition… you build an instance like you’re doing and tune it in enterprise edition, then you export from enterprise, and import into standard, the various plans and profiles you created. I have not done this part yet… only exported from enterprise and imported into another instance of enterprise. I suspect that when you tune it in enterprise and move to standard edition that the “self tuning” that would normally take place, does not continue to take place. It’s tuned at that point as well as the current conditions in enterprise allow… and it won’t ever* get worse; but, it also won’t ever get better. (*ever, barring major data or hardware changes since tuning, requiring re-tuning, since auto tuning won’t take place in SE).

That last bit in standard edition is just my suspicion… not having done it yet.

I believe it to be a valid business model to do this operation for multiple customers. Doing it for one may save tens to hundreds of thousands of dollars… doing it for many, just as scaleable. If you have a customer who really needs and can afford Oracle and is notably not using extra cost options: a priceless exercise! Repeatable for any Oracle user to some degree.

The currently entry point for EE with no options is $231,800 for the license…. easily double or triple if using options. Options, of course, do not migrate back to standard edition… so it’s a very select market as to who can benefit the most by tuning and running in standard edition. Many very likely would benefit from SQL Server or MySQL. Oracle is hard to justify to all but the heaviest users.

Hope some of this helps motivate you to learn the right tools. In time you will see I am right about the cloud, as well. Good luck with your EM installation my friend.

Leave a Reply

Your email address will not be published.