Good question. For me, it's an iterative process. If I get a false positive, I simply increase the threshold, and sometimes, turn it off completely (yes, you may quote me).Again, how do you tell whether or not the 20% variance was a problem or not ?
I'm not exactly sure what you mean here, but if you are saying that other metrics give the same alert, the answer is "yes"! When we get a real-world alert, several triggers usually fire together.And why couldn't the same problem appear without the variance occurring with it being swollen by the overall average activity of the database ?
I'm not sure that "redundant" is the best word to use here.You always need to check with some other metric or metrics regardless, hence the BCHR itself is redundant.
Yep, there are no absolutes in Oracle, just probabilities . . .If you get a false positive on one occasion, it might be a true positive the next time.
Exactly! And as Peter Sellers sez: "I like to watch" . . . . .Sometimes large management reports get run, sometimes they don't.
I don't think that DBA's have much of a choice there!Isn't letting the Users decide if response times are good, or not, a bad thing?
Whoa! Didn't you aleady concede that a change in BCHR indicates that something changed? Granted, it's not very useful, but taken alone, few metrics are. The BCHR is just one one tool in the toolbox . . . .Sometimes the radio can be off and it runs OK or bad.
My, what a strange thing to say . . . .Only it doesn't play David Bowie songs ...
Good analogy. That's sorta like the difference between first_rows and all_rows.That the driver would have been under the false impression that she was getting the best from her car would only have meant that she had got used to a substandard system; not that it was, all along, a good system.
My, sounds like a Silver Bullet. . . . Do you remember what you changed?After (admittedly quite simple) tuning, key queries suddenly ran a lot faster.
This topic does come up here from time to time. You can find one instance where the BCHR did point to a genuine problem Re: performance issue - buffer cache hit ratio below 50%. Basically the guy was running an actual system on the factory default settings for memory, values which would only have been considered right in the mid-nineties. A school boy error.Like I said, redundant.
I honestly don't look at the BCHR of any database I manage.
Bill Gates never was a developerActually, Bill co-wrote an early BASIC interpreter. The original storage media for it was paper punch tape (Computer Bowl, ca. 1990).