| Oracle FAQ | Your Portal to the Oracle Knowledge Grid | |
 |
 | |||
| Oracle FAQ | Your Portal to the Oracle Knowledge Grid | |
|
| |||
Home -> Community -> Usenet -> c.d.o.server -> Re: SQL*Net tuning - what's reasonable for response times
Well....since your statement "The multi-threaded server can't be
enabled on NT" doesn't specify which version of the Oracle database
either, someone (a newbie to Oracle) might think this means that there
is *no* version of Oracle that supports MTS on Windows NT. This is
false.
I have no source of amazement. I am just attempting to correct any misinformation.
On Wed, 19 Aug 1998 15:57:23 -0400, "Craig M. Wall" <cwall_at_petersons.com> wrote:
>The original post didn't specify what version of the database he's using
>and since most sites on NT are still on 7.x
>which CAN NOT support MTS
>my post is valid so...
>what's your source of amazement?
>
>
>Joel R. Kallman wrote in message
><35db1da0.27280397_at_newshost.us.oracle.com>...
>>Wow!!! So I guess the Oracle8 Database Assistant that comes with
>>Oracle8 8.0.4 on Windows NT is incorrect when it asks you if you want
>>to set up a database using Multi-Threaded Server.
>>
>>As well, pages 2-8 and B-5 from "Oracle8 Enterprise Edition Getting
>>Started" manual for Oracle8 8.0.4 on Windows NT probaly doesn't make
>>sense then when they refer to using MTS on Windows NT.
>>
>>Here is an excerpt from a support note within Oracle:
>>
>>"Note that Oracle8 will support 3GB of user address space on
>>NT 4.0 with Service Pack 3. Since Oracle8 will also support
>>the Multi-Threaded Server (MTS), the number of threads in the
>>Oracle process will not need to be so high and thus there will
>>not need to be such a large amount of virtual address space
>>for the threads' stacks."
>>
>>So....it is wholly inaccurate to say that Oracle MTS cannot be used on
>>Windows NT.
>>
>>
>>On Wed, 19 Aug 1998 09:11:03 -0400, "Craig M. Wall"
>><cwall_at_petersons.com> wrote:
>>
>>>Umh...
>>>The multi-threaded server can't be enabled on NT.
>>>Something about Oracle already being multi-threaded on that OS.
>>>So that won't help.
>>>But then again, is there really a problem?
>>>I've heard people complain of 20, 30 and even 60 seconds for
>>>SQL*Net connect times but not 4.
>>>Are all those 119 highwater users coming in over the same
>>>network card and IP address?
>>>If so, that may be the bottleneck.
>>>Investigate subnetting the network so that blocks of users
>>>come in over dedicated ethernet cards.
>>>NT 4.0 can't handle multi-homing, which is what
>>>will occur if more than one network card is connected
>>>to the same network (hence the subnetting approach).
>>>
>>>Regards,
>>>
>>>Craig M. Wall
>>>
>>>
>>>
>>>Joel R. Kallman wrote in message
>>><35dabacb.1979185_at_newshost.us.oracle.com>...
>>>>On Wed, 19 Aug 1998 21:06:20 -0700, Austin Durbin
>>>><adurbin_at_nsearthlink.net> wrote:
>>>>
>>>>>I have a third party application that needs to connect to the database
>>>>>over a local area network and run a faily simple query to obtain a
>>>>>customer balance - all in less then 6 seconds.
>>>>>
>>>>>Some performance measurements show that 4 seconds is being
>>>>>consumed by establishing the connection (login). On average the database
>>>>>has 50 to 70 active connections at any given time with a highwater mark
>>>>>of 119 connections.
>>>>>
>>>>>Everything is running on NT 4.0 (workstation and server) and the network
>>>>>protocol is TCP/IP on 100mb ethernet. The server is a dual pentium with
>>>>>1GB of memory.
>>>>>
>>>>>Question: does 4 seconds to login seem reasonable? I've never before
>>>>>given much though to tuning connect times - my focus has always been on
>>>>>tuning queries, memory usage, etc.. Frankly, 4 seconds strikes me as not
>>>>>bad.
>>>>>
>>>>
>>>>This may not necessarily be a SQL*Net/Net8 tuning issue, but rather an
>>>>issue of the overhead associated with establishing a session with the
>>>>database and then authentication of the user. On NT, you will be
>>>>creating a new thread for this user session, and there is
>>>>operating-system overhead in performing this operation.
>>>>
>>>>One solution for you to investigate is Oracle MTS (Multi-Threaded
>>>>Server). From the Oracle8 Concepts manual:
>>>>
>>>>"The multithreaded server configuration allows many user processes
>>>>to share very few server processes. The user processes connect to
>>>>a dispatcher background process, which routes client requests to
>>>>the next available shared server process.
>>>>
>>>>The advantage of the multithreaded server configuration is that
>>>>system overhead is reduced, increasing the number of users that
>>>>can be supported. A small number of shared server processes can
>>>>perform the same amount of processing as many dedicated
>>>>server processes, and the amount of memory required for each
>>>>user is relatively small."
>>>>
>>>>If there is some latency in the concurrent usage of your application,
>>>>you can see *dramatic* performance gains and reduced server-resource
>>>>consumption with the use of MTS.
>>>>
>>>>Hope this helps.
>>>>
>>>>>I have been unable to track down any information regarding tuning
>>>>>SQL*Net and would appreciate any recommentations on how I can reduce
>>>>>this 4 seconds unless the consensus is that 4 seconds is exceptionally
>>>>>good performance.
>>>>>
>>>>
>>>>Thanks!
>>>>
>>>>Joel
>>>>
>>>>Joel R. Kallman
>>>>Oracle Government, Education, & Health
>>>>Columbus, OH http://govt.us.oracle.com
>>>>jkallman@us.oracle.com http://www.oracle.com
>>>>
>>>>----
>>>>The statements and opinions expressed here are my own
>>>>and do not necessarily represent those of Oracle Corporation.
>>>
>>>
>>
>>Thanks!
>>
>>Joel
>>
>>Joel R. Kallman
>>Oracle Government, Education, & Health
>>Columbus, OH http://govt.us.oracle.com
>>jkallman@us.oracle.com http://www.oracle.com
>>
>>----
>>The statements and opinions expressed here are my own
>>and do not necessarily represent those of Oracle Corporation.
>
>
Thanks!
Joel
Joel R. Kallman Oracle Government, Education, & Health
Columbus, OH http://govt.us.oracle.com jkallman@us.oracle.com http://www.oracle.com
 |
 |