In light of yesterday’s post, here’s a short look at the early days of Ethernet.
Nowadays, we take Ethernet for granted. We plug a cable jack into the wall or a switch and we get the network. What’s to think about?
It didn’t start that way. In the 1960s and 1970s, networks were ad hoc hodgepodges of technologies with little rhyme and less reason. But then Robert “Bob” Metcalfe was asked to create a local-area network (LAN) for Xerox’s Palo Alto Research Center (PARC). His creation, Ethernet, changed everything.
On a related note, in one of the recent Xerox Alto restoration videos, two of the people who worked on the invention of Ethernet, Dave Boggs and Ron Crane, helped out fixing the Alto Ethernet card – carrying some very old-fashioned Ethernet equipment and telling some great stories from the early ’70s.
Sadly, Ron Crane passed away 19 June.
The availability of such a widespread ethernet standard proved extremely useful for the industry and I commend them for their work. However there are shortcomings that have held us back in hindsight, like the 1500 byte MTU. Engineers should have known this was going to be a future bottleneck and I feel it should have been anticipated.
1500B MTU makes sense for modem speeds and even 10mbps networks. But at 100mbps it makes less sense. And at 1gbps, 10gbps the overhead is alarming. Yet here we are using the same 1500 byte packets today. Modern equipment is often capable of handling jumbo frames but because they are neither standardized nor backwards compatible there’s no way to automatically determine a peer’s MTU on the network segment (path-mtu at the IP layer is different, and even that is held back by MTU issues at the ethernet layer).
The fact that the industry has evolved around hard coded 1500B packets has made jumbo frames impractical outside of specialized configurations. Ie when you turn it on by default things break, so vendors are forced to turn it off.
https://www.stsauver.com/~joe/jumbos/jumbo-frames.pdf
In their defense, they were probably just building ethernet as a solution to their own problems then and there, they didn’t conceive their technology becoming the basis for the world’s digital telecommunications networks decades into the future. This is probably the way most inventions go – they aren’t designed for the masses when they reach critical mass and outgrow their original purpose.
I have made a couple of Ethernet MACs, and for me, an important issue was processing at high speed very short frames back to back, 64bytes long (plus the 8 bytes Preamble/SoF and 10 bytes inter-frame gap, worst case is around 82 bytes time ;-). Once the headers are parsed, copying the data payload is easy.
We all encounter different problems…
One of the most forward-thinking aspect is the 48 bits MAC addresses. We will eventually run off of these addresses too, but designing such a gigantic address space in the late 70’s was really optimistic.
There is also some irony in the fact that Ethernet prevailed against competing standards partly because it supported a cheap shared wire with CSMA-CD at 10Mbps.
Since the BASE-T standards and the generalization of switches instead of hubs, this shared media with collisions concept is now abandoned.
Treza,
My post went off on a tangent
I had a network card that supported both 10mbps ethernet as well as 10baseT, but I only ever saw 10baseT used in an office setting. My LAN experience at home started with a 10mbps ethernet hub.
Yeah, that’s how I see it, too. Most people would rather stick with the “good ol’ familiar” IPv4 and just keep NAT’ing it to the grave.
I’ve worked in small/local to medium/regional, to huge/global companies and NONE of them want to do anything with IPv6. In fact, they disable IPv6 support on all the systems globally by default. Mostly because IPv4 “just works” as far as they’re concerned and IPv6 is still half-assed in most cases.
On the bright side — you can memorize, dictate IPv4 addresses pretty easily. No such luck with IPv6 😀
Heh, remember when switches were so expensive that people were actually building 10/100 hubs that had a single internal switch between 10 and 100? But they were still a hub. A 10 Mbps and a 100 Mbps hub.
zlynx,
At least there was no buffer bloat
https://www.bufferbloat.net/projects/
http://www.dslreports.com/faq/17883
Edited 2017-07-03 00:58 UTC
It’s still very useful for wifi, which is a modification of Ethernet.
You may as well say that the ancient Romans should have anticipated Ferraris and built their roads accordingly.
When ethernet was designed back in 1974 engineers were dealing with KB file sizes (or less). They could never have realistically anticipated the need for megabit speeds let alone gigabit speeds.
unclefester,
Well, it’s not so much about the file sizes, but rather the routing frequency that leads to the overhead/deterioration. For example we could be transferring a 15KB payload. Using 1500B MTU, with overhead this requires routers to route 11 or so packets to route the payload. With an 9216B MTU that goes down to 2 packets per payload. This is a significant gain that could have increased network capacity even for that time.
I suspect the reason they ignored this is probably because their job was to design ethernet for their own specific cases at PARC at the time rather than something that would become a world-wide standard. I cannot blame them over the scope of their job, however if they had set out to develop a world-wide networking standard, then I think engineers would have had an impetuous to make it more future-proof.
I’m not sure how old you are. Back in 1974 teletype terminals were still widely used. I doubt that PARC users needed to do anything much more demanding than sending text to printer or a bit of code to another terminal. Overheads were totally irrelevant.
unclefester,
(emphasis mine)
They were thinking about more than teletypes here. Unfortunately they don’t specifically outline their rational for the maximum transmission unit, but the document is an interesting read never the less.
Edited 2017-07-05 12:38 UTC
Lets not forget ALOHAnet … without it we wouldn’t have Ethernet, Wifi or GSM or Imarsat for that matter….
It basically set the stage for utilizing a random access shared channel which is something each of these technologies does and borrows directly from ALOHAnet.
Edited 2017-07-03 20:59 UTC
In the 90’s, I begun to learn about networking, and I actually remember it with fun. Twisted pair was not that widespread yet in my country at that time. We used coaxial with terminators all the way. The most fun thing was to do practical jokes with people. See… When ever computers try to communicate over coaxial, it is with a strong signal. So we used to have people changing the T-Plug, while the computers were running. To be extra shure, we had none of the machines grounded on the wall plug. Yeahh… Was fun to see them people get shocked.
Other than that. Fun memmories of the corridors of the dorm, litterly being a kind of snake farm, up against the walls, from all them coaxial cables. Was a time of DIY and experimentation and multiple computer platforms back in 1994/95. One of the best periodes in my life, untill they installed a professional setup of twisted pair and 10mbps hubs. All complete with wall outlets. I kind of felt like the magic was gone and it all had a kind of corporate feel to it.
Chaotic networking was fun, though we did mark each of our own cables with our name, so complete chaos was avoided. Yeah… A time when we were like a singulair chaotic yet democratic entity, regarding networking.
Before Co-ax we had a thick cable wwith markings to show where you could tap the cable. You clamped a tool around the cable and drilled through the outer shield to expose the inner core. Then you clamped the tap in place and prayed that the outer and inner bits were not touching.
As I said in the thread about Xerox, I still have the DEC issued Ethernet tapping tool and a short lenght of the thick co-ax cable. I also have a box of the later think Co-ax ‘T’ pieces and terminators plus a DEC switch.
I may donate it all to the NMoC when I’m up there next.
Amazingly named “Vampire taps”