On Sun, 14 Nov 2004 20:43:21 +0100,
(E-Mail Removed) wrote:
>I get confused with the different numbers concerning
>network and broadband throughput, and what they
>actually mean in real performance.
You're not alone.
>You know the bits
>and Bytes and how much gets lost or slowed down
>in the firewall / router etc.
>
>I am supposed to help a friend this coming tuesday
>set up a wireless network in his house. Since I have
>recently seen speeds listed up to 108 whatever, I
>told him that that is what we should shop for. Now
>however I can not find any router of that capacity
>that also has a print server, whichis a must.
http://www.netgear.com/products/details/FWG114P.php
If this is what you buy, make sure you get version 2. The original
version was a loser.
>His broadband currently gives 2 Mbs but will likely
>soon be upgraded to 8.
8Mbits/sec? Where's this?
>Obviously we would not
>want to invest in gear that will become a bottleneck
>in the near future.
Wireless hardware is NOT an investment. Everything you can buy today
will obsolete within a few years. That's not to say it won't be
useable or functional, but simply that advances in protocols,
security, acronyms, and features will make todays hardware somewhat
neanderthal. For example, improvements in 802.11i security will
probably require new hardware. 802.16 could replace everything. MIMO
(802.11n) technology could offer spectacular performance and
reliability improvements. Figure on a 2-3 year useful lifetime.
>What wifi speed corresponds to 8 Mbps?
>Is that 8 out of the 108?
>In that case 54 will be plenty for a long time.
108Mbits/sec is a proprietary kludge designed to squeeze more
performance out of 802.11g under ideal conditions. There are various
methods and chipsets, none of which are controlled by a standards
organization. You'll never see 108Mbits/sec thruput. Typicial is
less than half the connection speed. Thuput also varies with the type
of traffic. UDP streaming goes faster than TCP which requires
acknowledgements. Range is an issue as the radios will drop in
connection speed in an attempt to keep the error rate low. That will
cause a drop in thruput.
If you have a good strong signal, 802.11g will connect at 54Mbits/sec
and deliver 20-27Mbits/sec. That should be sufficient for anything
that you're doing. If your signal is less than spectacular, the
connection speed will be reduced along with the thruput. My guess(tm)
is that 802.11g will drop to about 25Mbits/sec connection, for a
thruput of about 8-10Mbits/sec, which should be adequate.
There's quite a bit on the topic of speed vs performance on:
http://www.tomsnetworking.com
in the "NeedtoKnows" section. It's a bit clumsy reading with all the
advertising and 15 part articles, but the content is worth the effort.
Chart of 108 and 54 802.11g performance vs range via a "walk test".
Note that thruput drops rapidly with distance, not matter what
technology is used, and that while 108 offers an initial speed
advantage, the performance drops rapidly to where it is the same as 54
802.11g.
http://www.tomsnetworking.com/Sectio...le60-page5.php
The dip in the middle was at 300ft, where performance was about
2Mbits/sec thruput. Extrapolating against the time scale, my guess is
that greater than 8Mbits/sec thruput will work to about 100ft.
It's gonna be interesting to see how 8Mbits/sec works with the default
TCP/IP RWIN (receive window) of 64KBytes. In theory, your maximum
latency (ping time) is:
Window Size / speed = latency
512Kbits / 8000Kbits/sec = 64 msec
If the latency greater than 64 msec, then the web server will stop and
wait for your TCP ack's to arrive before sending any more data. That
will slow things down. Windoze 2K and XP support largerer window
sizes but not all web servers do. You might wanna run a test:
http://www.dslreports.com/tweaks
to see where you stand when you get it running.
--
Jeff Liebermann
(E-Mail Removed)
150 Felker St #D
http://www.LearnByDestroying.com
Santa Cruz CA 95060 AE6KS 831-336-2558
Linear Mode
