Slow connection to Oracle 10
in [Oracle]
|
Prev: Database in quiesce mode
Next: download.oracle.com
From: Palooka on 8 Sep 2008 17:21 joel garry wrote: > On Sep 4, 7:48 pm, Michael Austin <maus...(a)firstdbasource.com> wrote: > >> Also, make sure you in your BIOS that you have hyperthreading turned >> OFF. As of 10.2.0.2, it was almost twice as fast turning this off. Not >> sure why this occurred, but it was consistent on more than 15 different >> x86 servers (HP, Dell, Compaq (it was ~2 years ago) NT2000/2003/XP) > > Some of the problems were due to java version issues, especially with > the EM agent. The more "sophisticated" your technology stack, the > bigger the argument to keep Oracle on the latest patch of the latest > version. Other times it's just as simple as Shakespeare showed. > > We can still laugh at Oracle for saying four years ago that "Oracle > will work just fine" on hyperthreading. (Note: 205089.1) > The thing which bothers me is that OP appears to have taken Shakespeare's solution (undoubtedly the correct one here) as a silver bullet, and edited his sqlnet.ora, without actually trying to understand what is going on (hint: it's going all over the Windoze network trying to find a database which is probably on the same box). IMHO, this attitude is unfortunate; it won't aid him in thinking through his next problem. Palooka
From: Shakespeare on 9 Sep 2008 01:49
"Palooka" <nobody(a)nowhere.com> schreef in bericht news:cHgxk.165201$W71.8946(a)newsfe12.ams2... > joel garry wrote: >> On Sep 4, 7:48 pm, Michael Austin <maus...(a)firstdbasource.com> wrote: >> >>> Also, make sure you in your BIOS that you have hyperthreading turned >>> OFF. As of 10.2.0.2, it was almost twice as fast turning this off. Not >>> sure why this occurred, but it was consistent on more than 15 different >>> x86 servers (HP, Dell, Compaq (it was ~2 years ago) NT2000/2003/XP) >> >> Some of the problems were due to java version issues, especially with >> the EM agent. The more "sophisticated" your technology stack, the >> bigger the argument to keep Oracle on the latest patch of the latest >> version. Other times it's just as simple as Shakespeare showed. >> >> We can still laugh at Oracle for saying four years ago that "Oracle >> will work just fine" on hyperthreading. (Note: 205089.1) >> > The thing which bothers me is that OP appears to have taken Shakespeare's > solution (undoubtedly the correct one here) as a silver bullet, and edited > his sqlnet.ora, without actually trying to understand what is going on > (hint: it's going all over the Windoze network trying to find a database > which is probably on the same box). > > IMHO, this attitude is unfortunate; it won't aid him in thinking through > his next problem. > > Palooka If you take a close look at the log files, you'll find the connection was waiting for NTS (15 secs twice). I did not suggest not using NTS as a solution, but merely offered him a test case yo show the delay WAS caused by NTS. Which makes more sense to me than just turn off HT because of some problem 2 years ago (which was a real problem by the way). Shakespeare |