Ethernet is Still Going Strong After 50 Years

I remember running Twin-axial cable for IBM token ring networks in a power plant. Dual coax and DB-15 wiring yuck.
Electrician shorted the connector end and the token got lost, or duplicated - took down the entire network and caused quite an uproar.

Ethernet I dislike that you can have problems at the physical layer, causing retries and retransmission yet not know this is going on. If the twisted pairs are damaged (stretched during cable pull, stepped on etc.), or the cable run is too long (some switches have weak drive), or the wrong cable is used (Cat. III lol).
Slower speeds but I have not seen any Ethernet hardware that tells you the retry counts are high.
 

Thread Starter

nsaspook

Joined Aug 27, 2009
12,796
Modern hardware on a good OS keeps good track of Ethernet errors.

ip -s -s link show dev eth0
2: eth0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc mq state UP mode DEFAULT group default qlen 1000
link/ether 98:b7:85:00:f1:1a brd ff:ff:ff:ff:ff:ff
RX: bytes packets errors dropped missed mcast
12837519799171 10430992610 2045 43 0 16082229
RX errors: length crc frame fifo overrun
2045 0 0 0 0
TX: bytes packets errors dropped carrier collsns
15340658002121 14113011742 0 0 0 0
TX errors: aborted fifo window heartbt transns
0 0 0 0 4

eth0: flags=4163<UP,BROADCAST,RUNNING,MULTICAST> mtu 1500
inet 10.1.1.172 netmask 255.255.255.0 broadcast 10.1.1.255
inet6 fe80::9ab7:85ff:fe00:f11a prefixlen 64 scopeid 0x20<link>
ether 98:b7:85:00:f1:1a txqueuelen 1000 (Ethernet)
RX packets 10395783218 bytes 12794267415523 (11.6 TiB)
RX errors 2045 dropped 43 overruns 0 frame 2045
TX packets 14076698347 bytes 15307582152182 (13.9 TiB)
TX errors 0 dropped 0 overruns 0 carrier 0 collisions 0

eth0:1: flags=4163<UP,BROADCAST,RUNNING,MULTICAST> mtu 1500
inet 10.1.1.85 netmask 255.255.255.0 broadcast 10.1.1.255
ether 98:b7:85:00:f1:1a txqueuelen 1000 (Ethernet)

eth1: flags=4163<UP,BROADCAST,RUNNING,MULTICAST> mtu 1500
inet xx.xx.xx.xx netmask 255.255.192.0 broadcast 50.53.63.255
inet6 fe80::9ab7:85ff:fe00:f11b prefixlen 64 scopeid 0x20<link>
ether 98:b7:85:00:f1:1b txqueuelen 1000 (Ethernet)
RX packets 3937286670 bytes 5124770249998 (4.6 TiB)
RX errors 0 dropped 0 overruns 0 frame 0
TX packets 1919322992 bytes 1172879658370 (1.0 TiB)
TX errors 0 dropped 0 overruns 0 carrier 0 collisions 0

enp1s0f0: flags=4163<UP,BROADCAST,RUNNING,MULTICAST> mtu 1500
inet 10.1.1.1 netmask 255.255.255.0 broadcast 10.1.1.255
inet6 fe80::9ab7:85ff:fe00:714a prefixlen 64 scopeid 0x20<link>
ether 98:b7:85:00:71:4a txqueuelen 1000 (Ethernet)
RX packets 18324619421 bytes 26636933283951 (24.2 TiB)
RX errors 4168 dropped 2224930 overruns 0 frame 4168
TX packets 1849832880 bytes 128981451515 (120.1 GiB)
TX errors 0 dropped 0 overruns 0 carrier 0 collisions 0

Most of the errors are from a 10 Gbps Ethernet backbone that's doing total image backups from all the servers to the backup server 10.1.1.1. That machine (a regular PC type instead of a enterprise type server) needs an upgrade to totally handle the I/O load but it's fine for now.
 
Last edited:

SamR

Joined Mar 19, 2019
5,019
Yup, we had coaxial Ethernet on our PDP-5 with proprietary workstations control system running 27 fractional distillation columns. Updated it to PDP-8 with DEC Alpha Workstations (before Compaq bought them) but still used coax ethernet. Along with HP mini-mainframe to coax ethernet for workstations in the main office and storeroom for business accounting, payroll, and inventory control. That all got replaced with ~250 HP Vectra desktops and I designed and had a contractor install a redundant fiber optic backbone to the entire plant facility to include messaging, Lotus123, and other desktop apps. Also had word processing for the secretarial staff to replace their standalone Wang word processors as well as a quite a few shared HP LaserJet printers. All on Cat5 to fiber and finally got rid of the plant coax. Still used a few ethernet extenders from the fiber boxes for some remotes but mostly simply Cat5 to fiber with a router rack in a central location for the fiber distribution. We had a lot of Cat I Div 1&2 explosive atmosphere areas throughout the plant that required conduit and sealing with explosive proof fittings for coax or Cat5 distribution so it wasn't until I put the fiber up that it became economical to "computerize" the entire plant except for in explosive proof control rooms that had to be stand-alone until I had the fiber run.
 
Last edited:

eetech00

Joined Jun 8, 2013
3,823
Yup, we had coaxial Ethernet on our PDP-5 with proprietary workstations control system running 27 fractional distillation columns. Updated it to PDP-8 with DEC Alpha Workstations (before Compaq bought them) but still used coax ethernet. Along with HP mini-mainframe to coax ethernet for workstations in the main office and storeroom for business accounting, payroll, and inventory control. That all got replaced with ~250 HP Vectra desktops and I designed and had a contractor install a redundant fiber optic backbone to the entire plant facility to include messaging, Lotus123, and other desktop apps. Also had word processing for the secretarial staff to replace their standalone Wang word processors as well as a quite a few shared HP LaserJet printers. All on Cat5 to fiber and finally got rid of the plant coax. Still used a few ethernet extenders from the fiber boxes for some remotes but mostly simply Cat5 to fiber with a router rack in a central location for the fiber distribution. We had a lot of Cat I Div 1&2 explosive atmosphere areas throughout the plant that required conduit and sealing with explosive proof fittings for coax or Cat5 distribution so it wasn't until I put the fiber up that it became economical to "computerize" the entire plant except for in explosive proof control rooms that had to be stand-alone until I had the fiber run.
You’re sure cat5? Might have been cat3(?)
 

eetech00

Joined Jun 8, 2013
3,823
Modern hardware on a good OS keeps good track of Ethernet errors.

ip -s -s link show dev eth0
2: eth0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc mq state UP mode DEFAULT group default qlen 1000
link/ether 98:b7:85:00:f1:1a brd ff:ff:ff:ff:ff:ff
RX: bytes packets errors dropped missed mcast
12837519799171 10430992610 2045 43 0 16082229
RX errors: length crc frame fifo overrun
2045 0 0 0 0
TX: bytes packets errors dropped carrier collsns
15340658002121 14113011742 0 0 0 0
TX errors: aborted fifo window heartbt transns
0 0 0 0 4

eth0: flags=4163<UP,BROADCAST,RUNNING,MULTICAST> mtu 1500
inet 10.1.1.172 netmask 255.255.255.0 broadcast 10.1.1.255
inet6 fe80::9ab7:85ff:fe00:f11a prefixlen 64 scopeid 0x20<link>
ether 98:b7:85:00:f1:1a txqueuelen 1000 (Ethernet)
RX packets 10395783218 bytes 12794267415523 (11.6 TiB)
RX errors 2045 dropped 43 overruns 0 frame 2045
TX packets 14076698347 bytes 15307582152182 (13.9 TiB)
TX errors 0 dropped 0 overruns 0 carrier 0 collisions 0

eth0:1: flags=4163<UP,BROADCAST,RUNNING,MULTICAST> mtu 1500
inet 10.1.1.85 netmask 255.255.255.0 broadcast 10.1.1.255
ether 98:b7:85:00:f1:1a txqueuelen 1000 (Ethernet)

eth1: flags=4163<UP,BROADCAST,RUNNING,MULTICAST> mtu 1500
inet xx.xx.xx.xx netmask 255.255.192.0 broadcast 50.53.63.255
inet6 fe80::9ab7:85ff:fe00:f11b prefixlen 64 scopeid 0x20<link>
ether 98:b7:85:00:f1:1b txqueuelen 1000 (Ethernet)
RX packets 3937286670 bytes 5124770249998 (4.6 TiB)
RX errors 0 dropped 0 overruns 0 frame 0
TX packets 1919322992 bytes 1172879658370 (1.0 TiB)
TX errors 0 dropped 0 overruns 0 carrier 0 collisions 0

enp1s0f0: flags=4163<UP,BROADCAST,RUNNING,MULTICAST> mtu 1500
inet 10.1.1.1 netmask 255.255.255.0 broadcast 10.1.1.255
inet6 fe80::9ab7:85ff:fe00:714a prefixlen 64 scopeid 0x20<link>
ether 98:b7:85:00:71:4a txqueuelen 1000 (Ethernet)
RX packets 18324619421 bytes 26636933283951 (24.2 TiB)
RX errors 4168 dropped 2224930 overruns 0 frame 4168
TX packets 1849832880 bytes 128981451515 (120.1 GiB)
TX errors 0 dropped 0 overruns 0 carrier 0 collisions 0

Most of the errors are from a 10 Gbps Ethernet backbone that's doing total image backups from all the servers to the backup server 10.1.1.1. That machine (a regular PC type instead of a enterprise type server) needs an upgrade to totally handle the I/O load but it's fine for now.
Turn on jumbo frame support. It’ll be faster.
 

Thread Starter

nsaspook

Joined Aug 27, 2009
12,796
Turn on jumbo frame support. It’ll be faster.
I know but there are other considerations on the network with a mixture of all sorts of prototype devices. It's plenty fast for what I do and less of a troubleshooting mess when things go bad.
 

djsfantasi

Joined Apr 11, 2010
9,151
Re things at the physical layer: plug one jack into another in an office. Boom goes the network!

A previous consultant had terminated a port with an RJ45 plug instead of a jack because “it would always be used for a small hub”. Well, the hub was discarded and the cable rolled up and stuck behind a printer.

All was good, until the printer stopped working. An office worker said, “This must be the problem.” And plugged it into a wall jack.

Boom went the network!

It was an event venue with a large wedding that afternoon. Fortunately, we tracked down the issue.
 

SamR

Joined Mar 19, 2019
5,019
You’re sure cat5? Might have been cat3(?)
Probably was, my memory is failing me these days... I'm not even sure it had a Cat classification anymore. I still have my crimpers and a bunch of plugs I used for making custom cables in a hurry. Also did some custom D shell RS-232 serial cables for engineering plotters and such. Quickly found out that serial cables maxed out ~50' or less even with using special 38400 baud rate drivers added onto early MS-DOS. Kept an old 300 baud acoustic coupler that you strapped the telephone handset into for years before I finally pitched it in the bin.
 
Last edited:

eetech00

Joined Jun 8, 2013
3,823
Probably was, my memory is failing me these days... I'm not even sure it had a Cat classification anymore. I still have my crimpers and a bunch of plugs I used for making custom cables in a hurry. Also did some custom D shell RS-232 serial cables for engineering plotters and such. Quickly found out that serial cables maxed out ~50' or less even with using special 38400 baud rate drivers added onto early MS-DOS. Kept an old 300 baud acoustic coupler that you strapped the telephone handset into for years before I finally pitched it in the bin.
Heh...Interesting.
If i remember correctly, Cat5 was released about the same time Fast Ethernet (10/100Mb) products were released, and "structured cable systems" became the "thing" to learn and implement.
In R&D we used PDP-11/73's and a DEC alpha server along with numerous Sparc workstations and Sparc Server. Also had HP9000 workstation with hard disk. The hard disk was about $14K and 5-10M space! (I don't remember disk space exactly, but it was tiny)
Started 10base and Cabletron hubs. Later changed to 10baseT and CDDI switches.
 

SamR

Joined Mar 19, 2019
5,019
I did the fiber as a project engineer but I was also assisting the IT dept. doing basic PC support. In the beginning we had HP field engineers coming to the plant weekly and they started leaving drives mostly as they were common failures and I'd get the call to and fix them but HP finally did away with their field engineers. Their Vectra hard drives also head drifted and had to have SpinRite rebuild the file structure before it completely failed which also kept me busy. I also found the solution to the Project Engineering dept. search for CAD. We had a quote from one of the major CAD systems for 2 engineering workstations and plotter for ~500k which put the project on the skids. I was also an ISA member (and officer) and at one of the trade shows we put on a guy had CAD running on a PC clone for cheap. So I put together a 1 seat and plotter system with tape backup for ~50k running VersaCAD and had it approved and installed which meant I also had to maintain it and train the drafters and expanded it to 3 seats and troubleshot problems as they arose and kept up with the upgrade and oncall maintenance contract. As an instrumentation Project Engineer I was also getting involved in automation upgrades installing Foxboro (and others) Distributed control systems. Problem was once installed there was no follow through by the operating areas for upgrades and maintenance. As thing evolved I was moved into the IT dept. (since they were the "computer" folks) and was in charge of specifying, installing, programing, backing up, and maintaining all of the various computer-based control systems on the plant as well as PLCs and drive controllers. HP had some kinky equipment. Their Vectras never could keep time and the failed drives I took home and upgraded my old PC clone with and they worked fine, just not in an HP box. Fun while it lasted...
 
Last edited:

Ian0

Joined Aug 7, 2020
9,503
I often wonder how many tonnes of copper are sitting unused in the blue and brown pairs of Cat/3/4/5/6 ethernet, and why it was decided in the first place to use 8-core cable when there was plenty of twin twisted-pair available as telephone cable.
 

eetech00

Joined Jun 8, 2013
3,823
I often wonder how many tonnes of copper are sitting unused in the blue and brown pairs of Cat/3/4/5/6 ethernet, and why it was decided in the first place to use 8-core cable when there was plenty of twin twisted-pair available as telephone cable.
Actually, the additional pairs are used in today's Gigabit ethernet (full duplex with flow control). Gig ethernet won't work correctly without them.

In the beginning 10BaseT ethernet was designed to work over Cat3 twisted pair cable, and many amateur network designers did use "telephone cable", like Silver Satin, for twisted pair networks. However, CRC errors caused by crosstalk and noise prevented the networks from operating reliably, especially with early implementations of Novell netware networks. Todays cable testers will fail a cable systems certification test if all pairs are not terminated end to end.
 
Last edited:

Thread Starter

nsaspook

Joined Aug 27, 2009
12,796
Heh...brings back memories, but I don't know why anyone would buy that nowadays.

Coax was a PITA to troubleshoot...
Very old industrial systems that are running just fine after 25 years and won't be replaced until the factory burns to the ground or is closed.
 
Top