31 Dec
2006
31 Dec
'06
10:38 a.m.
I'm beginning to wonder if it's time I upgraded my home network. The
NICs on most of the machines are 100BaseT, with the odd old machine
sporting a 10BaseT. But it's the hubs I wonder about--two old
National Semi 6-port Datamover 10BaseT units coupled together with a
10Base2 (coax) link. It's that coax that I dread replacing--it goes
between floors and snaking it through the walls was a real chore.
My DSL speed is 1.6Mbps, so I'm thinking that there's little reason
to upgrade the setup. Am I thinking correctly or am I likely to see
an improvement in internet access speed if I upgrade?
Cheers,
Chuck
31 Dec
31 Dec
11:12 a.m.
Chuck Guzis wrote:
I'm beginning to wonder if it's time I
upgraded my home network. The
NICs on most of the machines are 100BaseT, with the odd old machine
sporting a 10BaseT. But it's the hubs I wonder about--two old
National Semi 6-port Datamover 10BaseT units coupled together with a
10Base2 (coax) link. It's that coax that I dread replacing--it goes
between floors and snaking it through the walls was a real chore.
My DSL speed is 1.6Mbps, so I'm thinking that there's little reason
to upgrade the setup. Am I thinking correctly or am I likely to see
an improvement in internet access speed if I upgrade?
Cheers,
Chuck
I think you answered your own question.
Access to the outside world will not improve. Speeds between machines
will improve if you have switches and appropriate cabling.
I run 100Mbps for the more modern machines, all one a switch, and I
daisy chain an old hub (100 or 10, can't remember) for the slower
machines that only do 10Mbps.
Even with 5Mbps cable modem, my slowest connection inside is still
faster that my connection to the outside world.
Mike
12:54 p.m.
At 1:12 PM -0600 12/31/06, Michael B. Brutman wrote:
I run 100Mbps for the more modern machines, all one a
switch, and I
daisy chain an old hub (100 or 10, can't remember) for the slower
machines that only do 10Mbps.
I do similar, I have a 16-Port 100Base-T switch, that has most things
plugged into it, then I have a 8-Port hub with a 10Base2 connection
that lets me hook up a tiny thinnet segment (I only need it for my
DECserver 90L+ and Amiga 3000). In the near future I'll likely add a
small 1000Base-T switch, especially if I can ever find a 1000Base-T
card for my VMS Box, or SunBlade 1000. I wish I could find some dirt
cheap 1000Mbit Fibre to Copper media converters, Fibre 1000Mbit NIC's
are semi-cheap.
Even with 5Mbps cable modem, my slowest connection
inside is still
faster that my connection to the outside world.
You obviously don't have any really old systems on your network. I'm
still at 1.5Mbps, and have systems on my network that can't keep up
with that.
Zane
--
| Zane H. Healy | UNIX Systems Administrator |
| healyzh at aracnet.com (primary) | OpenVMS Enthusiast |
| MONK::HEALYZH (DECnet) | Classic Computer Collector |
+----------------------------------+----------------------------+
| Empire of the Petal Throne and Traveller Role Playing, |
| PDP-10 Emulation and Zane's Computer Museum. |
| http://www.aracnet.com/~healyzh/ |
1:22 p.m.
Zane H. Healy wrote:
You obviously don't have any really old systems on your network. I'm
still at 1.5Mbps, and have systems on my network that can't keep up
with that.
I can attest to that! My 11/70 which is running 10Mbps ethernet can
only ftp at ~25KB/s (and that's to a machine that's on the same switch!).
--
TTFN - Guy
2:49 p.m.
At 1:22 PM -0800 12/31/06, Guy Sotomayor wrote:
Zane H. Healy wrote:
Wasn't sure how well a Unibus system would handle things, but I know
that my VAXstation II/RC running VAX/VMS V5.5 and the CMU TCP/IP
stack was very slow, as is my PDP-11/73. Of course if I'm
remembering correctly the Q-Bus backplane is only 3Mb/sec.
I still remember how much of a speed difference I saw going from
running Linux on a 486DX/33 with an ISA NIC, to a Pentium 133 with a
PCI NIC.
Zane
--
| Zane H. Healy | UNIX Systems Administrator |
| healyzh at aracnet.com (primary) | OpenVMS Enthusiast |
| MONK::HEALYZH (DECnet) | Classic Computer Collector |
+----------------------------------+----------------------------+
| Empire of the Petal Throne and Traveller Role Playing, |
| PDP-10 Emulation and Zane's Computer Museum. |
| http://www.aracnet.com/~healyzh/ |
You obviously don't have any really old systems on your network.
I'm still at 1.5Mbps, and have systems on my network that can't
keep up with that.
I can attest to that! My 11/70 which is running 10Mbps ethernet
can
only ftp at ~25KB/s (and that's to a machine that's on the same
switch!).
4:47 p.m.
Well, here's why I asked.
The system sitting off of the second hub seems to get about 10% lower
throughput (same O/S, same NIC) off the 'net than any system sitting
off the first hub (all much slower systems).
Could the second hub and the translation/turnaround over the coax
link between the two be the culprit?
Just trying to figure out why a 2 GHz box is getting consistently
beaten by a 225 MHz one.
Cheers,
Chuck
6:22 p.m.
It was thus said that the Great Chuck Guzis once stated:
Well, here's why I asked.
The system sitting off of the second hub seems to get about 10% lower
throughput (same O/S, same NIC) off the 'net than any system sitting
off the first hub (all much slower systems).
Could the second hub and the translation/turnaround over the coax
link between the two be the culprit?
Just trying to figure out why a 2 GHz box is getting consistently
beaten by a 225 MHz one.
Did you try putting the 2GHz system on the first hub?
Other problems including speed and duplex mismatches between the 2GHz
system and the hub it's plugged into (I've run into that problem a few times
at work) or perhaps a marginal cable (between the computer and hub, or hub
to hub).
-spc (Just some ideas ... )
6:32 p.m.
At 9:22 PM -0500 12/31/06, Sean Conner wrote:
Other problems including speed and duplex mismatches
between the 2GHz
system and the hub it's plugged into (I've run into that problem a few times
at work) or perhaps a marginal cable (between the computer and hub, or hub
to hub).
I've had nightmarish problems over the years with bad ports on
switches. I spent quite a bit of time trying to figure out why I was
having so much trouble with my PDP-11 running RSTS/E and DECnet/E,
and it turned out to be the one switch. Another time I was getting
crazy slow speeds between a 2Ghz P4 running XP and a dual 2Ghz Mac.
Zane
--
| Zane H. Healy | UNIX Systems Administrator |
| healyzh at aracnet.com (primary) | OpenVMS Enthusiast |
| MONK::HEALYZH (DECnet) | Classic Computer Collector |
+----------------------------------+----------------------------+
| Empire of the Petal Throne and Traveller Role Playing, |
| PDP-10 Emulation and Zane's Computer Museum. |
| http://www.aracnet.com/~healyzh/ |
6:29 p.m.
You wrote...
Just trying to figure out why a 2 GHz box is getting
consistently
beaten by a 225 MHz one.
This sounds like the classic problem of a duplex mismatch between the device
and the hub. A lot of folks don't realize it, but when you mix
10mb/HDX/100mb/FDX devices in a single network - you would THINK that the
devices would autonegotiate identical settings on both ends. C'mon, they
even SAY they autonegotiate on the box, and they DO have an 'auto' setting,
right? Fact is, they get it wrong better than 80% of the time. Even two
devices from the same manufacturer usually don't autonegotiate right.
Pretend that "autonegotiation" doesn't exist. Set both ends manually to
identical settings. If one
Some will say - this can't be... I plug them in and they work, and the right
speed lights come on! And they pass traffic JUST fine. Don't trust this.
FDX/HDX mismatches can create very odd looking wierdness. Like a connection
that SEEMS to work, but is actually working very poorly.
Jay
9:27 p.m.
On 31 Dec 2006 at 20:29, Jay West wrote:
Some will say - this can't be... I plug them in
and they work, and the right
speed lights come on! And they pass traffic JUST fine. Don't trust this.
FDX/HDX mismatches can create very odd looking wierdness. Like a connection
that SEEMS to work, but is actually working very poorly.
That may well be the problem, thanks! I'll check into it.
Cheers,
Chuck
1 Jan
1 Jan
12:27 a.m.
On 31 Dec 2006 at 20:29, Jay West wrote:
This sounds like the classic problem of a duplex
mismatch between the device
and the hub. A lot of folks don't realize it, but when you mix
10mb/HDX/100mb/FDX devices in a single network - you would THINK that the
devices would autonegotiate identical settings on both ends.
I took some older Socket 7 boards with both ISA and PCI slots with
CPUs ranging from an Intel 166 (no MMX) to a K6 PR300. Tried about 5
different PCI 100BaseT NICs and only a single generic NE2000 10BaseT
ISA card. In every case, the ISA card was faster than the PCI cards
(talking to a 10BaseT hub) when doing an A-B comparison in the same
box. OSes were Win98 and Win2K. I suppose I should throw in some
Linux just to be fair, but I didn't do it this round. To be fair, I
used different cables and ports on the hub.
I think you've hit the nail on the head, Jay--at least it seems that
way. I've got a few more tests to do before I'm sure. To say that
I'm surprised is an understatement.
Cheers,
Chuck
9:31 a.m.
At 08:29 PM 12/31/2006 -0600, you wrote:
You wrote...
Yeah, I can
attest. We used to have one particular run of 3Com cards that
negotiated the wrong duplex about 25% of the time when used with one
particular kind of 10/100 blade in a Cisco Catalyst 5000 series switch. The
symptoms were that the connection would work ok, and ping tests seemed to
indicate all was well, but as you loaded the connection down a little, it
would start having errors and retries, until the retries overwhelmed any
real work being done.
Any time someone reported a good connection that would hang, timeout, and
quit with errors whenever you tried to copy a file over it (larger than a
few k) I know what the problem was.
A firmware upgrade fixed some of it. Policy fixed the rest: Thou shalt set
the speed and duplex we tell you to on your PC, or we'll disable your port.
No Auto/Auto permitted. Better gear and Corporate Standard NICs eliminated
the problem after a while.
[Computing] The use of COBOL cripples the mind; its teaching should,
therefore, be regarded as a criminal offence. --E.W.Dijkstra, 18th
June 1975.
--... ...-- -.. . -. ----. --.- --.- -...
tpeters at nospam.mixcom.com (remove "nospam") N9QQB (amateur radio)
"HEY YOU" (loud shouting) WEB ADDRESS http//www.mixweb.com/tpeters
43? 7' 17.2" N by 88? 6' 28.9" W, Elevation 815', Grid Square
EN53wc
WAN/LAN/Telcom Analyst, Tech Writer, MCP, CCNA, Registered Linux User 385531
Just trying to figure out why a 2 GHz box is
getting consistently
beaten by a 225 MHz one.
This sounds like the classic problem of a duplex mismatch between the
device and the hub. A lot of folks don't realize it, but when you mix
10mb/HDX/100mb/FDX devices in a single network - you would THINK that the
devices would autonegotiate identical settings on both ends. C'mon, they
even SAY they autonegotiate on the box, and they DO have an 'auto'
setting, right? Fact is, they get it wrong better than 80% of the time.
Even two devices from the same manufacturer usually don't autonegotiate
right. Pretend that "autonegotiation" doesn't exist. Set both ends
manually to identical settings. If one
Some will say - this can't be... I plug them in and they work, and the
right speed lights come on! And they pass traffic JUST fine. Don't trust
this. FDX/HDX mismatches can create very odd looking wierdness. Like a
connection that SEEMS to work, but is actually working very poorly. 
2 Jan
2 Jan
12:50 p.m.
Tom Peters wrote:
A firmware upgrade fixed some of it. Policy fixed the
rest: Thou shalt
set the speed and duplex we tell you to on your PC, or we'll disable
your port. No Auto/Auto permitted. Better gear and Corporate Standard
NICs eliminated the problem after a while.
Alvarion stuff is bad for this, particularly when used in conjunction
with Cisco switches. The most recent Cisco firmware just makes the
problem worse, because if you specify the speed and duplex settings on
the switch, it will fall back to auto/auto if it loses its connection.
Like, for instance, when you reboot the wireless bridge...
Gordon (still gearing up to bomb Tel Aviv until Alvarion sort out their
stupid software)
1 Jan
1 Jan
12:25 a.m.
Chuck Guzis wrote:
Could the second hub and the translation/turnaround
over the coax
link between the two be the culprit?
Only if you're running at half duplex and/or have some sort of collision
nightmare.
Just trying to figure out why a 2 GHz box is getting
consistently
beaten by a 225 MHz one.
Boot knoppix (linux on a CD that runs completely off of the CD) and run
"netstat -in" to see if the 2Ghz box is suffering from something obvious.
--
Jim Leonard (trixter at oldskool.org) http://www.oldskool.org/
Help our electronic games project: http://www.mobygames.com/
Or check out some trippy MindCandy at http://www.mindcandydvd.com/
A child borne of the home computer wars: http://trixter.wordpress.com/
8:56 p.m.
Guy Sotomayor wrote:
Zane H. Healy wrote:
That might be because of limited disk bandwidth. The machine itself is
capable of talking faster than that.
Peace... Sridhar
You obviously don't have any really old systems on your network. I'm
still at 1.5Mbps, and have systems on my network that can't keep up
with that.
I can attest to that! My 11/70 which is running 10Mbps ethernet can
only ftp at ~25KB/s (and that's to a machine that's on the same switch!).
31 Dec
31 Dec
1:33 p.m.
Zane H. Healy wrote:
The context was the networking gear, not the machines. I think my PCjr
is fairly slow.
In the case of a really old system like that, it doesn't matter what the
network is - the network will never be the problem, unless I use bongos
as the transport mechanism.
Mike
Even with
5Mbps cable modem, my slowest connection inside is still
faster that my connection to the outside world.
You obviously don't have any really old systems on your network. I'm
still at 1.5Mbps, and have systems on my network that can't keep up with
that.
Zane
10:02 p.m.
On Sun, 2006-12-31 at 15:33 -0600, Michael B. Brutman wrote:
In the case of a really old system like that, it
doesn't matter what the
network is - the network will never be the problem, unless I use bongos
as the transport mechanism.
Keep it in historical context... Bongos were a significant speed
increase over smoke signals.
Also, I note that you COULD have a slow network (10baseT) have an
adverse effect upon your 1.5 Mbit (IIRC) broadband connection. That
could happen if the network is jammed with traffic, and the collision
rate is such that the packets from the broadband can't get through in
timely fashion. In that case, increasing the network speed of the local
machines would reduce the time they spent occupying the network. Most
times, it won't matter, granted, but it COULD slow you down.
Peace,
Warren E. Wolfe
wizard at voyager.net
10:19 p.m.
Warren Wolfe wrote:
On Sun, 2006-12-31 at 15:33 -0600, Michael B. Brutman
wrote:
The bongo reference was tongue in cheek - I was alluding to this:
The Bongo Project: TCP/IP via Primitive Communication
http://eagle.auc.ca/%7edreid/index.html
The point being that my slowest machines will never be able to saturate
a 10Mbps Ethernet, or cause the collisions at the rate you allude to.
My PCjr can do at best 40KB/sec over the Ethernet, quite a bit less if
disk I/O is involved.
But you're welcome to try - the Jr is up and running. :-) (See my
previous email about TCP/IP Testing help)
Mike
In the case of a really old system like that, it
doesn't matter what the
network is - the network will never be the problem, unless I use bongos
as the transport mechanism.
Keep it in historical context... Bongos were a significant speed
increase over smoke signals.
Also, I note that you COULD have a slow network (10baseT) have an
adverse effect upon your 1.5 Mbit (IIRC) broadband connection. That
could happen if the network is jammed with traffic, and the collision
rate is such that the packets from the broadband can't get through in
timely fashion. In that case, increasing the network speed of the local
machines would reduce the time they spent occupying the network. Most
times, it won't matter, granted, but it COULD slow you down.
10:38 p.m.
> In the case of a really old system like that, it
doesn't matter what the
> network is - the network will never be the problem, unless I use bongos
> as the transport mechanism.
On Mon, 1 Jan 2007, Warren Wolfe wrote:
Keep it in historical context... Bongos were a
significant speed
increase over smoke signals.
??
~1100 feet per second v 186,000 miles per second
But bongos could operate without line of sight,
and with or without ambient light.
Neither seems to be supported by current MICROS~1 software.
So, reliability is likely to be higher than modern NICs.
1 Jan
1 Jan
9:15 a.m.
On Sun, 2006-12-31 at 22:38 -0800, Fred Cisin wrote:
On Mon, 1 Jan 2007, Warren Wolfe wrote:
Oh, SURE, the PROPAGATION speed is higher for smoke signals, the
delay comes in filling the "cache" with smoke to let it go, and
different interpretations of various clouds of smoke, necessitating
error retries.
Keep it in historical context... Bongos were
a significant speed
increase over smoke signals.
??
~1100 feet per second v 186,000 miles per second But bongos could operate without line of sight,
and with or without ambient light.
Yes, an attribute not seen again in the commercial market until the
advent of SneakerNet.
Neither seems to be supported by current MICROS~1
software.
So, reliability is likely to be higher than modern NICs.
The EXISTENCE of software from MICROS~1 does not, in itself, degrade
a NIC -- one has to actually RUN the software to mess things up. My NIC
has only run Linux drivers, and has functioned perfectly.
Peace,
Warren E. Wolfe
wizard at voyager.net
12:22 a.m.
Chuck Guzis wrote:
Am I thinking correctly or am I likely to see
an improvement in internet access speed if I upgrade?
You will see 0% improvement in internet access speed. However, moving
your entire internal network to 100mbps makes things like network
storage a convenience, and 1000mbps makes it transparent.
Half a decade ago I upgraded everything to 100mbps and it became
possible for me to edit video over the network (DV, which is 3.125MB/s,
easily within the realm of 100mbps). That really helped my workflow, as
I could put all my crap in one place but edit any project from any
machine I wanted.
--
Jim Leonard (trixter at oldskool.org) http://www.oldskool.org/
Help our electronic games project: http://www.mobygames.com/
Or check out some trippy MindCandy at http://www.mindcandydvd.com/
A child borne of the home computer wars: http://trixter.wordpress.com/
2 Jan
2 Jan
10:08 a.m.
At 12:38 PM 12/31/2006, Chuck Guzis wrote:
My DSL speed is 1.6Mbps, so I'm thinking that
there's little reason
to upgrade the setup. Am I thinking correctly or am I likely to see
an improvement in internet access speed if I upgrade?
I vote for the upgrade, just on principle. Pulling new wire
rejuvenates the spirit. I think the blood sacrifice improves
the transfer times, too. Here is some old advice on this topic,
at the bottom of the page:
http://www.ipprimer.com/ethernet.cfm
If I were to summarize Jay's advice as further justification for
an upgrade, it would be "if you don't measure, you don't know."
You've measured throughput and discovered some slow spots, but
did you also measure latency, and would improved latency make
your contemporary computers happier on the Internet?
- John
2:08 p.m.
On 12/31/06, Chuck Guzis <cclist at sydex.com> wrote:
It's that coax that
I dread replacing--it goes
between floors and snaking it through the walls was a real chore.
Depending upon how many bends and whether it is fastened to the walls,
it might be a fairly easy replacement. Cat 6 cable is close enough to
the the same diameter as coax, so it should fit in the same run. I've
done it by splicing the Cat 6 to the coax. Getting a splice that
isn't of larger diameter than the coax is the tough part. The
tightest splice I managed to do involved a needle, fishing line, glue,
and heat shrink tubing, but that case was extreme due to unreachable
metal grommets in the middle of the run. Once the splice is done and
you are convinced it is strong enough, just pull it through.
My DSL speed is 1.6Mbps, so I'm thinking that there's little reason
to upgrade the setup. Am I thinking correctly or am I likely to see
an improvement in internet access speed if I upgrade?
You're not going to see a speed increase to the outside, but you might
see a speed increase between machines on the local net. It's useful
if you have modern machines that share files on both floors/
On the rack next to me is a 10/100bT switch, a 10/100bT hub, a 10bT
hub with coax uplink, an ethernet-localtalk interface, and 54 Mbps
wireless access point. The cable modem and another 10/100bT switch
are downstairs in the garage. Now if I could find a way to get
Arcnet into the mix.....
6330
days inactive
6332
days old
22 comments
14 participants
participants (14)
-
cclist＠sydex.com -
cisin＠xenosoft.com -
ggs＠shiresoft.com -
gordon＠gjcp.net -
healyzh＠aracnet.com -
jfoust＠threedee.com -
jwest＠classiccmp.org -
korpela＠ssl.berkeley.edu -
mbbrutman-cctalk＠brutman.com -
ploopster＠gmail.com -
spc＠conman.org -
tpeters＠mixcom.com -
trixter＠oldskool.org -
wizard＠voyager.net