Re: [Q] Web Speed OCI/Oraperl

From: Vernon Keenan <vkeenan_at_us.oracle.com>
Date: 1996/05/04
Message-ID: <vkeenan-0405961252310001_at_ppp11.lanminds.com>#1/1


In article <199605031620.MAA01028_at_astor.mediabridge.com>, thayer_at_astor.mediabridge.com wrote:

> Our oraperl CGIs (under sloaris) are awefully slow and we're
> investigating ways to speed them up. What's not yet clear is where
> the majority of time is spent in these short-lived processes.
> The candidates are :
>
> a) waiting to connect to the web server (apache)
> a: possible kernel network tweaking
> b) starting the cgi itself
> a: modify the httpd to incorporate the cgi directly
> find a way to reduce the oracle binary bloat
> c) connecting to the database
> a: preconnect (maybe a pool of cgis waiting to run)
> d) ora_open
> a: is there a way to pre-parse statements and/or cursors??
> e) queries: perhaps our sql is really just aweful
> a: reorganize table.
>
> If you have experience with these problems please let me know.
>
> /charles

I think you'll find the performance bottlenecks are steps (B) (C). That's why we enhanced Oracle WebServer 2.0 to use a API to access a continuously running connection daemon that's always connected to Oracle. Same thing with Netscape Enterprise Server 2.0 with LiveWire Pro.

When Rob McCool invented CGI at NCSA everybody knew it was a hack, including Rob. That's why there are APIs for Netscape, Apache, Microsoft IIS and Oracle servers. If you're hard core you could write your own Pro*C-based interface to Oracle using the Apache API. Otherwise, I'd recommend taking a look at the Oracle WebServer at the URL below.

Vern

-- 
+----------------------------------------------------------+
| Vernon Keenan                 | 500 Oracle Parkway       |
| Product Marketing Manager     | Web/Workgroup Division   |
| Oracle Corporation            | Redwood Shores, CA 94065 |
| E-mail: vkeenan_at_us.oracle.com | Phone: (415) 506-4651    |
|                 WWW: http://www.oracle.com/              |
+----------------------------------------------------------+
Received on Sat May 04 1996 - 00:00:00 CEST

Original text of this message