Oracle FAQ Your Portal to the Oracle Knowledge Grid
HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US
 

Home -> Community -> Usenet -> c.d.o.server -> Re: Quirky performance differences/problems between connections made by Unix client versus Windows Client

Re: Quirky performance differences/problems between connections made by Unix client versus Windows Client

From: Michael Blakeley <mike_at_blakeley.com>
Date: Fri, 11 Jan 2002 21:05:37 GMT
Message-ID: <mike-4A5C9E.13055711012002@news.attbi.com>


In article <4c08015.0201110305.1af52498_at_posting.google.com>,  m_patte_at_yahoo.com (Mark PAtterson) wrote:

> I'm running Oracle 8.1.6 and 8.1.7 databases on a Solaris 8 server and
> am experiencing quirky performance differences between Solaris based
> clients and Windows based clients.
>
> In a nutshell clients running on Solaris (including sqlplus running on
> the database server) seem to freeze up every now and again when
> attempting to connect or process a select statement etc. I don't even
> need to be doing data modifications (hence no redo info or locking
> problems).
>
> However, Windows based clients never experience same symptoms and I've
> even found that executing a simple select statement via the Windows
> client (SQL*Plus) actually free up the frozen Solaris client. This
> "cure" can be applied and works consistently!
>
> Any ideas out there please?
>
> Additional Info: Using MTS for both Windows and Solaris based clients.

I recently saw some similar problems in a JDBC application. We weren't able to resolve them entirely, but here are some ideas.

Historically, the Windows TCP/IP implementation has been buggy enough to cause performance problems between (for example) Solaris web servers and Windows clilents (see
<http://www.sun.com/sun-on-net/performance/tcp.slowstart.html>). The number of hops on your network might exacerbate this problem.

Also, take a look at 'netstat -s' from both sides. On Solaris, it's handy to redirect the output into a 'before' file and an 'after' file, and diff them. See any retransmits, errors, or other suspicious-looking stats?

--
Michael Blakeley       mike@blakeley.com     <http://www.blakeley.com/>
            Performance Analysis for Internet Technologies
Received on Fri Jan 11 2002 - 15:05:37 CST

Original text of this message

HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US