Where do you want Ethernet ? Everywhere apparently.
I’m at a conference about “Cloud Ethernet Forum” and subsidiary operation of the Metro Ethernet Forum that seems to want to invent a new Ethernet. The fundamental premise of the is that customers need/want/demand/must have more Ethernet in the WAN.
Why ? Because Ethernet is better for data centre managers.
Immediate response ? The SDH/Sonet and DWDM I have been using for the last decade is working really well and getting better. Why would people even remotely think this is a good idea ?
The Cloud Ethernet Forum(CEF) has it’s primary goal to develop methods, standards & technologies that support large scale Ethernet WAN deployments and especially Data Centre Interconnects for Layer 2 extension.
Thus, Ethernet from the data centre to the provider, provider-provider exchanges, Ethernet exchange points (multi-provider) and then in the final mile. In short, CEF wants to enable Ethernet as an end-to-end WAN technology that service providers can scale with visibility, specifically suited for use in connecting data centres and covered in Unicorn glitter (I made that last one up).
What is the Cloud Ethernet Forum ?
CEF provides a unique consensus-building framework for industry stakeholders to collectively develop solutions that address technical challenges such as VLAN scaling, layer 2 performance and resilience across very large domains and consolidating storage network technologies onto Ethernet.
Decoding that: * “consensus-building” = anyone who wants to speak up regardless of competency, expertise or experience. * “layer 2 performance and resilience” = someone who clearly hasn’t experienced Ethernet in the WAN because resilient Ethernet hasn’t been achieved in 20 years. * “consolidating storage network technologies onto Ethernet” = errr, storage people hate the IP protocol so lets pander to their idiot requests.
What do I need from a WAN ?
Bandwidth. It’s all about bandwidth. Whatever I have, it’s not enough. Of course, service providers don’t want to sell bandwidth, there is no “enhanced value” in that. Services are where the profits are !!! Providers always have to deliver services.
Apparently bandwidth is not a service. Or rapid circuit delivery. Or reliable technical support. But I digress.
But service providers connect cloud providers. And apparently cloud providers want to use private WAN connections to connect to enterprise customers. In other words, just like frame relay and ATM, customers are crying out for ways to have private networks to connect to their public cloud providers.
I don’t subscribe to this belief system either. People THINK they want dedicated circuits because that’s what they already have. Or because private circuits are secure – they aren’t
To use Ethernet I need to have visibility of my WAN end to end – because Ethernet doesn’t have that capability.
I guess that if Cloud Ethernet Forum can introduce a standard that provides end-to-end traffic visibility like Frame Relay then that would be an improvement. They are also promoting a vision of Ethernet peering exchanges where carriers and hosting providers can transparently interconnect Ethernet connects in a similar way to Frame Relay/ATM, BGP or MPLS Carrier to Carrier. Except, of course, this time it would work.
What does a WAN use Ethernet for ?
There is Ethernet as a LAN technology and Ethernet as a WAN Bearer. So I guess the question is whether Ethernet makes a good WAN technology. If we consider what the MEF standards have developed around technologies like PBB and QinQ then there are some scaling opportunities
You can still route IP over the Ethernet. Or use MPLS or some other encapsulation. The CEF boosters see Ethernet as a WAN protocol just like Frame Relay, ATM or DWDM. Instead
Dumb Ideas Supported and Welcome
Using Ethernet in the WAN support is a proverbial shotgun, loaded and pointed at your foot with your finger on the trigger. All sort of dumb ideas are supported and the following were mentioned to me as justification:
- What about Storage over Ethernet ?
- What about offering Data Centre Interconnection as a native service ?
- Hybrid Cloud Services so that customer doesn’t have to change their IP address scheme ?
- How about terminating 100 000 Ethernet VLANs at the edge of the managed hosting data centre with a consistent end-to-end VLAN allocation strategy for each customer ?
Ethernet can be force for good but, inevitably, people will choose the dumb option every time. Because they can.
Ethernet as Frame Relay
I’ve done a quick peruse of the “Cloud Ethernet” standards and basically they have taken most of the ideas of Frame Relay control plane and added concepts like end-to-end circuit status using ILMI, PVC and SVC creation with PPP-like protocols.
The IEEE Stumbling Block
The IEEE is the custodian of the Ethernet standard. The same standards body that delivers Ethernet standards like 10GBaseSR at 100metres because that is what we always do instead of 30metres which would have been 80% cheaper.
The IEEE is “just so good”¡ that the Metro Ethernet Forum was established to create new standards without the hassle of the IEEE. Yay. So now we have technologies like Provider Bridges and QinQ. Except the originators of those standards didn’t think big and they don’t really scale operationally.
So the CEF is integrating with MEF. Oh, and the ONF, IEEE, IETF and I don’t know, everyone. Except that no one has heard of CEF.
Security Not Considered
CEF has no interest in addressing link security or end-to-end data integrity and state that this is a customer issue to address. Second, there is no interest is addressing carrier integrity as highlighted by Snowden/NSA.
Congratulations to the people involved for not caring about security. I’m told that customers aren’t “asking for security features” but I’m quite confident that they are but it’s too hard for Metro Ethernet Forum to solve this because
- Their lack of technical chops
- Because security adds to the cost of technology. Short term-ism is rife in the discussions that I had with the CEF/MEF folks.
The EtherealMind View
I’m a big Ethernet fan. Really. Just because it’s the cheapest protocol that has lots of shortcomings at scale is not a reason to dislike it. But we should also recognise that Ethernet everywhere is not a good idea. If building Ethernet in the WAN is good idea because you want to use those cheap Ethernet switches and silicon, then develop a “Ethernet Frame Relay” protocol. Or just use Software Defined Networking.
It’s very clear that the MEF has a serious lack of clear thinking that is driven by carrier’s reacting to the market instead of leading it. Customers want more bandwidth and security. Ethernet doesn’t do either of these things.
I can’t shake the idea that the Metro Ethernet Forum is living in cuckoo land. Just because you can, doesn’t mean you should.
I attended a press event in Nice, France where I was briefed by the Cloud Ethernet Forum as a subsidiary operation of the Metro Ethernet Forum. Clearly, I was not overly impressed by the content but the event was an independent press event. My full disclosure statement is here