Kbase 16329: Buffer Hits calculation overflows in promon
Autor |
  Progress Software Corporation - Progress |
Acesso |
  Público |
Publicação |
  5/10/1998 |
|
Buffer Hits calculation overflows in promon
There is a problem with the Buffer Hits calculation in promon.
If customers report that they normally see a good number for
Buffer Hits (in the high 80's or 90's) that suddenly drops
dramatically (to maybe 10 or 20 percent or less) for no apparent
reason, they are likely seeing this problem.
The problem occurs when another number in promon, "Logical Reads",
gets very large. You can find this number in promon by going to
R&D -> 2. Activity Displays -> 3. Buffer Cache. When Logical reads
gets to about 45000000 or so, the buffer hits calculation will
overflow. If this happens, an approximate calculation for buffer
hits can be gotten from the other numbers on the Buffer Cache screen:
Buffer Hits ~= ((Logical reads - O/S reads) * 100)/Logical reads
This calculation should be accurate to within a few percent.
The real solution to the problem is to use the sampling features of
promon. To get a good measurement of Buffer Hits, you should pick a
busy time of day for your system. Get Buffer Hits for the same time
each day and for the same sampling interval. To set the sampling
interval, go to R&D-> 5. Adjust Monitor Options-> 3. Monitor sampling
interval (some versions of promon show this as option 7., but it's
really 3.) Set the interval to a reasonable length of time (600
seconds seems to work well).
Now go back to the Buffer Cache screen. Choose S to sample. The
screen will freeze for the sampling interval you selected. When the
interval is up, you will get numbers that reflect the interval you
selected. By repeating this at the same time of day and for the same
interval, you will get numbers that are useful for comparison. And
because sampling resets all counters, you will avoid the overflow
problem in the Buffer Hits calculation.
JHD (11-01-96)
Progress Software Technical Support Note # 16329