[Xitami] content/type headers
R O
xitami@lists.xitami.org
Mon, 18 Apr 2005 08:50:31 -0700
Thanks. That tip came from 14 years as a network engineer, which is my
primary technical expertise. Web Hosting and web design are side hobbies of
mine that keeps me somewhat current with servers, windows and other system
specific trends. However, server performance is just as inportant as
network performance. If you are aware of both you increase your overall
optimization. No server is completely independent, nor is any network.
NIC settings, duplex mismatches, L2 switch parameters, L3 switch/router
protocols - all affect network performance. You can have the fastest server
in the world sitting on an improperly configured network (or vise versa) and
see the worst response times. It would be like putting 87 octane into an
Ferrari that requires 103 octane. She's gonna choke and sputter and
possible stall out.
If I can think of any other tips, along these lines, I will put them out
there. This tip just happened to be related someone elses topic.
Russel P. Olinger
(310) 466-3399
rolinger1@hotmail.com
>From: "Johannes Prost" <xitami@pastryworld.ca>
>Reply-To: xitami@lists.xitami.org
>To: xitami@lists.xitami.org
>Subject: Re: [Xitami] content/type headers
>Date: Sat, 16 Apr 2005 11:49:19 -0600
>
>want to second that.
>
>Johannes
>
> > Hello Russel,
> >
> > It was a great tip, how can I find more tips such as this. Thanks.
> >
> > Inanc
> >
> > On 4/16/05, R O <rolinger1@hotmail.com> wrote:
> > >
> > >
> > > That would make sense. 1500 would be the max size packet anyone would
>want
> > > to create. Most router circuits are set to 1545 MTU (max transmission
> > > unit). Anything larger than that requires the router to split the
>packet
> > > giving the first 1500 to FRAME 1 and giving the reamainder to a new
>FRAME.
> > > That new packet though can't be used by any other data. so if your
>server
> > > packets are 1700, the first 1500 bytes goes to the first frame, then
>200
> > > bytes goes to the second frame, but the second frame is then filled up
> > > with
> > > 1300 bytes of nothing simply to fill the frame - it has to happen this
>way
> > > because a FRAME can only carry on TCP header. Larger packets causes
>the
> > > router to take a heavier processing load and slightly greater
>bandwidth
> > > load.
> > >
> > > Apply that logic to smaller server packets, say 850bytes per packet.
>Then
> > > every frame for that TCP transmission will be 850bits of data and 600
>bits
> > > of nothing. This causes the router to waste nearly 45% of its
>bandwidth
> > > (for that particular tcp session). If your packets match the MTU of
>the
> > > router then no framing is wasted, the router doesn't have to parse
>packets
> > > which decreases processing/reassembly time, and excess frames do not
>have
> > > to
> > > be sent.
> > >
> > > Russel P. Olinger
> > > (310) 466-3399
> > > rolinger1@hotmail.com
> > >
> > > >From: Marc Doigny <marc@verfaillie.com>
> > > >Reply-To: xitami@lists.xitami.org
> > > >To: xitami@lists.xitami.org
> > > >Subject: Re: [Xitami] content/type headers
> > > >Date: Fri, 15 Apr 2005 18:27:59 +0200
> > > >
> > > >At 16:46 15/04/05, you wrote:
> > > >>Hello Marc,
> > > >>
> > > >>very handy to control ones own server and to be able to check the
>server
> > > >>software of a
> > > >>site without the need of using Netcraft. What is the reason for the
> > > limit
> > > >>of bytes ?
> > > >>Otherwise one would be able to check how the full page will look
>like (
> > > >>development).
> > > >>
> > > >>Is there the possibilty to share this ? If so, what would be needed
>?
> > > >>
> > > >>Johannes
> > > >
> > > >I created the application just to check if my home server was
>responding
> > > to
> > > >internet requests, where it doesn't matter that 1500 bytes or more
>are
> > > >displayed.
> > > >1500 bytes seemed a good compromis between speed and size; the data
>must
> > > be
> > > >downloaded before it can be send to the client. Xitami is set to kill
> > > >applications running more than 10 seconds, and in some cases
>connecting
> > > and
> > > >downloading large files could not be done in 10 seconds.
> > > >Two connections are open during the whole transaction.
> > > >
> > > >
> > > >There are two versions available:
> > > >
> > > >http://web.verfaillie.com/cgi-bin/gap?www.verfaillie.com (displaying
>the
> > > >content as text)
> > > >
> > > >http://afr_upd.verfaillie.be:20800/cgi-bin/gap?www.verfaillie.com
>(trying
> > > >to display the content as html)
> > > >
> > > >The source and executable are disponible at
> > > >http://web.verfaillie.com/cgi/gap.bas
> > > >http://web.verfaillie.com/cgi/gap.exe
> > > >It just runs as a cgi-bin executable.
> > > >
> > >
> >-------------------------------------------------------------------------
> > > >--::Mailserver Verfaillie Bauwens::-- contact :
>postmaster@verfaillie.com
> > > >Alle mails gecontroleerd op virussen. Virusdefinities dagelijks
>aangepast
> > > >----------------------- http://www.verfaillie.com
>-----------------------
> > > >
> > > >
> > > >--
> > > >Xitami Users Mailing List -- For Xitami support
> > > >To unsubscribe: http://lists.xitami.org/mailman/listinfo/xitami
> > >
> > > --
> > > Xitami Users Mailing List -- For Xitami support
> > > To unsubscribe: http://lists.xitami.org/mailman/listinfo/xitami
> > >
> >
> >
> >
> > --
> > ---
> > life is nearly what you believe
> >
>
>
>
>--
>Xitami Users Mailing List -- For Xitami support
>To unsubscribe: http://lists.xitami.org/mailman/listinfo/xitami