[sci.electronics] LAN speed?????

wallwey@boulder.Colorado.EDU (WALLWEY DEAN WILLIAM) (04/13/90)

Does anybody know the realistic speed limit of data transfer that can
reliably be sent over RG-58 (or even the 75-ohm RG-59).  I think
thin Ethernet uses the RG-58.  Is Ethernet close to the limits of this
type of cable or if a different standard was adopted, could the performance
upscaled a magnitude or two?

[Personally I think LANS should be able to go lot faster than then
the 10 megabit/sec. standard Ethernet---Anybody actually know?  I'm
basing this guess on the fact that I'm able to receive 40 channles of
real time videa over RG-59(just like RG-58 only different impedence) via
local cable.   If each station requires about 6MHz of bandwidth alone,
than ethernet should be no where near the limits of the RG-58 cable.

Am I messed up or on the right track?
Please comment, easy on the flames.....

	Dean Wallwey

arnief@tekgvs.LABS.TEK.COM (Arnie Frisch) (04/13/90)

In article <19603@boulder.Colorado.EDU> wallwey@snoopy.Colorado.EDU (WALLWEY DEAN WILLIAM) writes:
>
>Does anybody know the realistic speed limit of data transfer that can
>reliably be sent over RG-58 (or even the 75-ohm RG-59).
>
>[Personally I think LANS should be able to go lot faster than then
>the 10 megabit/sec. standard Ethernet---Anybody actually know?  I'm
>basing this guess on the fact that I'm able to receive 40 channles of
>real time videa over RG-59(just like RG-58 only different impedence) via
>local cable.   If each station requires about 6MHz of bandwidth alone,
>than ethernet should be no where near the limits of the RG-58 cable.
>
RG-58 has 1.4 db of loss per 100 ft at 10 megaHertz.
The standard ethernet cable has less loss than this, but long runs
obviously degrade its signal quality.

The thru-put of ethernet is much lower than 10 megabits per second
because of protocol and SW overhead.  An excellent system can run at 1
megabit persecond.  Most workstations run at 100 to 300 kbits per
second.

Arnold Frisch
Tektronix Laboratories

BILLW@MATHOM.CISCO.COM (William Westfield) (04/13/90)

>[Personally I think LANS should be able to go lot faster than then
>the 10 megabit/sec. standard Ethernet---Anybody actually know?  I'm
>basing this guess on the fact that I'm able to receive 40 channles of
>real time videa over RG-59(just like RG-58 only different impedence) via
>local cable.   If each station requires about 6MHz of bandwidth alone,
>than ethernet should be no where near the limits of the RG-58 cable.

yes, ethernet speed is more limited by other factors (like everyone on
the cable haveing to hear the same packets and detect collisions reliably)
than by inherent limitations in the cable.  Similar cable used in broadband
networks has a total bandwidth much larger than 10 Mb, but the technology
is also much more complex (requiring repeaters, head ends, and other forms
of access control).  Things like "ultranet" run multiple 100Mbits over
coax, though not for the distances ethernet can travel over.


>The thru-put of ethernet is much lower than 10 megabits per second
>because of protocol and SW overhead.  An excellent system can run at 1
>megabit persecond.  Most workstations run at 100 to 300 kbits per second.

This is very obsolete information.  ANY workstation ought to be able to
get over 1 Mb these days, or be laughed at.  Good workstations get more
like 3 Mbps, and highly tuned implementations get better than 8 Mbps.
(yes, using real TCP, with checksumming, and even though there WERE
ethernet collisions during the data transfer.)

Even moderate speed PC/AT clones do better than 100 kbits most of the time.

Bill Westfield
cisco Systems.

henry@utzoo.uucp (Henry Spencer) (04/13/90)

In article <7301@tekgvs.LABS.TEK.COM> arnief@tekgvs.LABS.TEK.COM (Arnie Frisch) writes:
>The thru-put of ethernet is much lower than 10 megabits per second
>because of protocol and SW overhead.  An excellent system can run at 1
>megabit persecond.  Most workstations run at 100 to 300 kbits per
>second.

I guess that depends on whose workstations one is using :-), and on who's
writing the software.  Van Jacobson can drive an Ethernet to saturation
sending data from one Sun-3/50 to another with TCP/IP.  The net data
rate is 6-7 Mb/s, as I fuzzily recall, after you subtract the bit times
needed for headers etc. for TCP, IP, and Ethernet itself.

This is obviously not done with a stock SunOS TCP/IP :-), but it's *not*
a specially-cooked benchmarking-only implementation.  That man really
knows how to speed up protocol processing.

It used to be common wisdom that 1 Mb/s was about tops per host.  It's
not true any more.
-- 
With features like this,      |     Henry Spencer at U of Toronto Zoology
who needs bugs?               | uunet!attcan!utzoo!henry henry@zoo.toronto.edu

whit@milton.acs.washington.edu (John Whitmore) (04/13/90)

	There are busses (Futurebus+) that can transfer circa 250Mbytes/sec,
and fiber optic (FDDI) proposed hardware for higher throughput,
as well as RF modems (the campus here uses some) that can
get virtually the same throughput that a cable TV operation 
needs (about .5 Gbit/sec).  None of these, however, could
replace Ethernet.  Excepting Futurebus, they aren't supported 
through the ISO levels down to the message level (I think I got 
the right level; corrections gratefully accepted), so they are NOT networks.  
Futurebus has a length limitation (fifty feet?) that keeps it
from being a simple competitor.
	The Ethernet packet-delivery and collision sensing are
its beneficial features; if you need more throughput, offload 
devices from the trunk Ethernet and use a bridge to connect
them (i.e. faster Ethernet is equivalent to using more Ethernets).
There are many advantages to this approach (a blabby node doesn't
tie up everyone's Ethernet, an axe can't stop all traffic
for a mile radius, a local data source can be kept private from
listeners on other Ethernets), and the hardware manufacturers
like it, too! :->

 I am known for my brilliance,              John Whitmore
by those who do not know me well.

greene@venice.SEDD.TRW.COM (John Greene) (04/13/90)

In article <19603@boulder.Colorado.EDU> wallwey@snoopy.Colorado.EDU (WALLWEY DEAN WILLIAM) writes:
>
>Does anybody know the realistic speed limit of data transfer that can
>reliably be sent over RG-58 (or even the 75-ohm RG-59).  I think
>thin Ethernet uses the RG-58.  Is Ethernet close to the limits of this
>type of cable or if a different standard was adopted, could the performance
>upscaled a magnitude or two?

The limitation is not the data rate but the protocol.  The higher the data
rate, the 'shorter' the packet is and the end-to-end length of the network
needs to get shorter.  This is based on the idea that a transmitted packet
has to reach the last node on the network before the originating node
finishes transmitting.  One way to increase the data rate would be to use
a token bus protocol that would eliminate the need for collision detection.
This is how ChipCom gets past the distance limitation of 10broad36 with their
Marathon Bridge.  (get it?  Marathon, goes the distance?) 

Naturally, the higher the data rate the more susceptible it will be to 
flaws in the network structure.  I don't think you could get away with using
'T' connectors at 100 Mbit/sec data rates.  

In a way, a higher data rate is already being achieved because you can have
several 10broad39 units on a single RG-59 network.  They are modulating an
RF carrier at a 10 Mbit/sec rate so with three 10broad36 channels you network
data rate in 30 Mbit/sec.  It's just not between the same two nodes.

I have proposed designs in the past for 100 Mbit/sec transceivers/modems but
could never generate enough interest to have it launched as a project.  :-(
Mainly because it would utilize a token bus at a non-standard rate and they
didn't want to put money into an activty that was non-standard compliant.
"We think it's a great idea! but....."


-- 
John E. Greene    "People are just like frankfurters....You have to decide
                   if you're going to be a hot dog or just another wiener" DLR
TRW Systems Engineering and Development Division
INTERNET: greene@venice.sedd.TRW.COM  USENET: ..trwrb!venice!greene

phil@pepsi.amd.com (Phil Ngai) (04/14/90)

In article <19603@boulder.Colorado.EDU> wallwey@snoopy.Colorado.EDU (WALLWEY DEAN WILLIAM) writes:
|
|Does anybody know the realistic speed limit of data transfer that can
|reliably be sent over RG-58 (or even the 75-ohm RG-59).  I think

My company has a product called TAXI which I like to think of as
a 100,000,000 bit per second UART. The actual "baud per second"
is 125 Mbps because it is self-clocking, but the user only gets
100 Mbps.

I'm not in that group so I really don't know too much about it.
Perhaps someone else at AMD can speak on the subject.

--
Phil Ngai, phil@amd.com		{uunet,decwrl,ucbvax}!amdcad!phil
The War on Drugs is the modern day Inquisition.