Bob Metcalfe said I looked too old to possibly be me. Pot.
Two weeks ago I was at the Computer History Museum to help observe the 40th anniversary of Ethernet. It was literally 40 years to the day since Bob Metcalfe drew his first sketches of what became the world’s dominant OSI Level 2 network technology. It was a fun and festive day that ended, believe it or not, with dancing. But some of the celebration didn’t make sense to me. Or, rather, it seemed to me that important parts of the discussion were missing.
When, for example, did 802.11 WiFi become a part of 802.3 Ethernet? That claim was made over and over during the day and helped power the argument that Ethernet is today a $100 billion business. Yeah, right.
Parts of that argument are true. Certainly Ethernet is dancing on the grave of Arcnet and Token Ring. When Howard Charney of Cisco said that Ethernet accounts for 98 percent of the network interfaces shipping now from Cisco (if you pretend that WiFi is Ethernet) it had a world of meaning for the rise of Metro Ethernet and the decline of DS1 (T-1) and DS3 (T-3) connections — connections that have been Cisco’s profit centers until just now.
So this claim of all-Ethernet, all the time, is a big deal and has vast implications for Cisco’s future business and that of its competitors, but it doesn’t mean Ethernet is now the only game in town.
Cisco has made a fortune on traditional telephone circuits: (DS1 and DS3 interfaces) and CCIEs (Certified Cisco Internetwork Experts — don’t call them engineers) have made a good living keeping those clunky router interfaces running. Yet it hasn’t made sense to me for years that I could get 15+ megabits-per-second for around $30 through my cable or DSL modem when businesses were paying 15 times as much for 1.5-megabit T-1s that offered no scalability and problematic provisioning. That’s what appears to be finally changing and with it will change the required skill sets of CCIEs.
If the new network is really Ethernet end-to-end (LAN-to-WAN) then only a subset of CCIE training will be soon needed, right?
Not so fast.
Here’s another, slightly different, view of where Ethernet is going thanks to discussions with friends and CCIE’s like my friend George Morton who are living in those network trenches. Where one of the themes of the Museum event for example, seemed to be SONet is dead, it might be just as correct to say SONET is Ethernet.
I know, Mr. Smartypants, that SONET is Layer 1 while Ethernet is Layer 2, but Bob Metcalfe, himself, spoke at the event of Ethernet as a “SONET-killer,” yet I don’t see SONET disappearing soon where fiber is already installed.
What really stands in the way of Metro Ethernet are DOCSIS 3.0, (170Mbps/110Mbps) and 3.1, (10Gbps/2.5Gbps) — data technologies from your cable company.
The network in the next few years will see further LAN migration away from Ethernet toward WiFi with the NSA wireless security Suite-B and the new 1.3Gbps 802.11ac and ae WiFi protocols. The best way to think about this transition is there’s no way to easily do wired Ethernet with mobile devices so as the client mix changes, WiFi will naturally come to dominate.
From an operational cost perspective why pay cable costs of $100 to $200 for each copper Ethernet port, then buy a switch for each very expensive copper cable?
Cisco sees this change and as a result recently bought Meraki for cloud management of WiFi services.
You can see this trend toward wireless already with the number of WiFi-only products that connect to the LAN in your home like those cheaper Roku boxes that don’t even include an RJ45 port.
So here’s the real state of Ethernet at 40:
Data Center networking will be 10/40/100GbE.
The Metropolitan Area Network (MAN) will be Ethernet over Multiprotocol Label Switching (MPLS — a combination of Layer 2 and 3 technologies) or Virtual Private LAN Service (VPLS) which is layer 2 — essentially wide area Ethernet. But the Layer 1 protocols will be SONET, and DOCSIS, with DOCSIS eventually winning because of lower cost of deployment, (cable is cheaper to deploy versus fiber to the building).
The way DOCSIS wins for MAN and WAN access is because of lower cost. Here are some numbers from my own recent network shopping trip:
Service Monthly Rate
DS1 with MPLS and router $550
DS3 with MPLS and router $2,250
DOCSIS Private MAN 2Mbps: $310
DOCSIS Private MAN 50Mbps: $1,265
These DOCSIS services, by the way, were both Comcast Business Class Ethernet Network Service Premium with latency guaranteed less than 12ms, jitter less than 2ms, and packet loss less than 0.001 percent.
The market supremacy of Ethernet in the WAN and MAN, then, is not so much about the protocol but about operating costs. Looking at it another way, it’s all just marketing.
Even Metro-Ethernet ISPs have a hard time matching DOCSIS prices because of the cost of build-out. Telephone companies never intended fiber to be in the local loop, only twisted pair. Ethernet will shift from the LAN protocol to become the MAN/WAN protocol for edge delivery.
Who is going to deliver the bits is the question and in the end price will determine that. Remember ATM? It wasn’t bad, just too expensive.
How, if at all, will the sharing of federal spectrum (http://cis471.blogspot.com/2013/06/executive-order-directs-sharing-of.html) and new wireless technology impact connectivity of LANs?
We used to call ATM ‘A Terrible Mistake’ – not because it wasn’t a cool technology, but because it was a) too damn complicated for it’s own good and b) when technology pundits saw it not getting enough traction outside of the infrastructure space to be profitable, they tried pushing it to the wiring closet and then to the desktop, ergo problem a).
I bought, futzed with and cussed and more Xylan switches and their clunky CLI and external provisioning boxes than I can remember.
http://gcn.com/articles/1998/11/23/navy-will-install-xylan-atm-switches-aboard-four-ship-groups.aspx
I see a lot of offices getting 10 Mbps Metro E with a combination of of Internet connectivity, WAN to the datacenter where their main servers are, and SIP trunking to their phone systems. Once there is fiber to the building, metroE pricing is pretty competitive.
Well, what _is_ today’s Ethernet anyway? Initially, Ethernet was a technology centered around a cable, piercing taps, Manchester coded bits, a frame structure, addressing schema, parity check and most importantly an access control protocol (CSMA/CD). Today, if we talk about Ethernet, only the frame structure, addressing schema and parity check remain. There is a bunch of new cable and coding standards and a lot of implementations of Ethernet frames over something else (SDH/Sonet, DOCSIS, WiFi.). So today’s Ethernet is basically a frame structure.
Although most Ethernet frames are transported over full duplex links, we still have Collision Counters in our interfaces, dutifully reported by network monitoring software.
Well Bob, when you go to the museum you usually expect to see old artefacts, and Mr. Metcalfe may have thought you were on display! Only kidding of course. He looks a lot older than you do. Good article, thanks.
http://news.cnet.com/8301-1023_3-57589353-93/is-cable-holding-back-superfast-broadband-adoption-on-purpose/
What about Infinera’s 100G PIC system, coming to Metro in companies such as TW Telecom?
Bell Canada is deploying fibre to the home. Here in downtown Toronto, I can’t get DOCSIS 3.0. Heck, I’m not getting the full potential of DOCSIS 1.0, and Roger’s isn’t moving to fix it. So Cringely, your numbers look backwards from over here. Home user, 50Mbps symmetric for 50/month on fibre, or 5Mbps (with 750kbps upload) for 75. My Bell price has no transfer caps too.
In the end, copper will never keep up with fibre speed-wise, and you only need to run fibre once, backhoe-fade not-withstanding.
This is just irrelevant semantics based on a outdated term. “Ethernet” has evolved into a generic term for efficient user-end interfaces. Evolution of languange, that’s all. The common man doesn’t care for the specific details of the physical layer (nor should he need to) and that’s fine. Also, he doesn’t wan’t know what IEEE 802.XX his bits run over in general. What you are talking about is just WAN. The whole power of telecommunication systems is that the physical layer is obfuscated away, and the easier this becomes the better. So yes, WiFi is “Ethernet”, WAN is “Ethernet” and as this 3/4/5G malarky becomes more confusing, so will the cellphone.
“…WiFi is “Ethernet”, WAN is “Ethernet” and as this 3/4/5G…” I’m afraid we are a long way from that ideal. Consumers still need to be aware of the huge differences and tradeoffs in speed, cost, and convenience. Perhaps some of us will live to see the day when everything is as good as 100 mbps Ethernet or better.
The biggest problem with this discussion is most residents will never see 50/75/100Mbps, let alone 25, because we live in rural America where .we are at the mercy of the expensive telco-sat tv/telco-cable duopoly. Rural is not Brooklyn or even a DC suburb. It is driving through 50 miles of corn and soybeans to reach the nearest city. It is where 1.5/256 is common. It’s where DSL over copper is called POTS and 10/1 doesn’t create a warm and fuzzy feeling when the phone does not work. Cost is double. 10/1 with phone is $90/mo. 10/2 cable no phone is $70/mo. 15/5 cable no phone is $175. It is no telco cash cow when 12,000 residents compete with 30 million acres of corn and beans for a reliable and affordable Internet. If FTTH ever came to the home, it would be so expensive, the only option many of us would have is the 56k dial-up. I still have my modem – which may or may not even work where I live. To do anything wireless means a two year contract for a phone I don’t want. There is no easy solution either phone company wants to provide at any low cost.