Re: Performance debugging tips

Hi. i hope you understand my english, im from Ecuador, South America.

Oracle has a lot of techniques to make faster DML operations (Data manipulation language), those could be selects, inserts, updates or deletes; one of this techeniques is that any time you do a "select" a lot of records go from the hard disk to memory and the most recently used keep there, so with this, oracle ensure that the most used records are in memory, and the next time a select ask for it, oracle just go to the memory and dont go to the disk, and it make a faster query.

Maybe what happens with your debug is that you try a lot of times your proccess and any time you try there exists diferents records in memory because your other tries. And this make the diference.

You can make tunning to the oracle db, by telling oracle which are those records than should be in memory always, to make faster your proccess.

gfsolis.

stacy_friedman_at_hotmail.com ha escrito:

> Hi everyone,
>
> I'm trying to debug an inconsistent performance problem with an app
> that runs against Oracle. The general gist is that it takes in many
> smallish records, does a bit of analysis on each, and returns a result
> based on some DB operations -- an insert, then some queries, for each
> record. Most records are handled on a <0.1 second timeframe, but every
> once in a while we see one jump into the 3+ second range.
>
> The inconsistent part is, of course, that *which* record takes >3
> seconds is always different. If we do multiple trials of the same 100K
> records, we get different "bad" ones each time. That tells me that
> it's probably *not* the data, which is the same each time, nor the
> application, but something in the DB design. I'm trying to debug this
> -- guessing that it could be some sort of internal hash collision or
> something, but where else should I be looking? I'm not 100% familiar
> with all of the DB-level tools that come with Oracle for this sort of
> debugging, so feel free to be verbose. :)
>
> Thanks in advance for any help,
>
> Stacy
Received on Sun Oct 02 2005 - 02:20:33 CDT