Printable View
Hi Folks
I need to prove it to management that runing a statspack report without the setting the timed_statistics parameter equal to true is pretty useless.Does somebody have a document that provides me arguments which can strengthen my case.
regards
Hrishy
well if you dont set timed_statistics to true in the time columns inside statspack report you get all zeros, isnt that a good proof?
tell them to unhook the speedomoter from their car, tell em it's the same thing, you'll never know how fast your going.
Mt Hanky I smell an international Smack a Manager Day coming on. Are we done celebrating Smack a developer day?
Hi
Well thats obivious but since the wait count values are not zero and hence management wants me to go ahead..and do the analysis
regards
Hrishy
Quote:
Originally posted by stmontgo
tell them to unhook the speedomoter from their car, tell em it's the same thing, you'll never know how fast your going.
Mr Hanky I smell an international Smack a Manager Day coming on. Are we done celebrating Smack a developer day?
Tell them they will have to wait for their report. Tell them they will only have to wait once BUT refuse to say how long.Quote:
Originally posted by hrishy
Well thats obivious but since the wait count values are not zero and hence management wants me to go ahead..and do the analysis
Good one ;)Quote:
Originally posted by DaPi
Tell them they will have to wait for their report. Tell them they will only have to wait once BUT refuse to say how long.
hrishy,
show them the following output.
1 select event, total_Waits, time_waited, time_waited/total_Waits ratio from v$system_event
2* order by 4
SQL> /
EVENT TOTAL_WAITS TIME_WAITED RATIO
---------------------------------------- ----------- ----------- ---------
Null event 1 0 .00
process startup 6 0 .00
row cache lock 6 0 .00
instance state change 2 0 .00
library cache load lock 1 0 .00
SQL*Net message to client 43428308 6674 .00
control file sequential read 20316552 72123 .00
SQL*Net more data from client 494200 3896 .01
file identify 1640 14 .01
SQL*Net more data to client 967456 15492 .02
refresh controlfile command 3384027 93470 .03
db file sequential read 14781104 434406 .03
LGWR wait for redo copy 2967 96 .03
file open 61494 2506 .04
buffer busy waits 58427 5080 .09
db file scattered read 50092015 5373244 .11
library cache pin 7 2 .29
log file sequential read 97778 28377 .29
direct path read 1308378 432086 .33
latch free 21245 9365 .44
log file parallel write 609272 358121 .59
local write wait 6 7 1.17
db file parallel read 25 30 1.20
rdbms ipc reply 416 540 1.30
direct path write 1102171 1768816 1.60
log file sync 254377 494635 1.94
control file parallel write 455872 1224963 2.69
log file single write 818 2204 2.69
library cache lock 2 13 6.50
db file parallel write 62358 407415 6.53
single-task message 106 1423 13.42
SQL*Net message from client 43428348 669025305 15.41
log buffer space 933 16085 17.24
SQL*Net break/reset to client 299 7643 25.56
log file switch completion 293 11706 39.95
rdbms ipc message 6172950 283683546 45.96
pmon timer 463261 57165842 123.40
enqueue 913 128797 141.07
smon timer 4674 57132985 12223.57
Tell them that some events wait for longer and others wait for shorter period. so, the no of waits do not give exact picture.
Hi All
now this means that there were 56,100 waits maybe in 100,000cs or maybe in 15secsCode:Top 5 Wait Events
~~~~~~~~~~~~~~~~~ Wait % Total
Event Waits Time (cs) Wt Time
---------------------------- ------ ------------ -------
db file scattered read 56,100 0 .00
db file sequential read 38,457 0 .00
SQL*Net message to dblink 19,606 0 .00
SQL*Net message from dblink 19,600 0 .00
direct path read 11,006 0 .00
if former the wait event maybe significant if later then they can be ignored
can i base my arguments like above
regards
Hrishy
Base your argument by confusing them...the fact is you know more...and then smack them!!
or ask them on what knowledge do they base this question...and then smack them!!
trueQuote:
Originally posted by hrishy
event maybe significant if later...
maybe yes, maybe no. Any time you have the top 5 events and they're not idle events, it probably warrents investigation. However, like others have pointed out, until you know the time, the number of waits is insignificant.Quote:
...then they can be ignored