Share what you know with millions of people
Focus is the best place to turn what you know into remarkable content
40GigE...then 100GigE...when will the Ethernet era finally end?
Best Answer
You've started an interesting discussion, and ended it with promotion of your product.
But where is the question?
This doesn't seem to belong in the Q&A area. Maybe writing a white paper would be more appropriate, and educational for those that come to Focus to learn and share.
Andrew: Had Ethernet been efficient, Cisco would have never been born. That's when Ethernet became switched Ethernet. Nobody draws a distinction between the two anymore because very few networks can run on a four port hub without connectivity to the outside world.
The National Science Foundation thinks this issue is important enough to devote $30MM to academic research via GENI and FIND grants, and if you look at any of the RFPs from NASA, DARPA or the NSF, they all ask for network architectures that have nothing to do with the current state of the art.
Todd: Point taken. I wish I had a product to promote, but we're still at proof of concept. However, the question from the title of the discussion remains (re-phrased), "Does anyone have a solution that will enable society to migrate to a future Internet architecture?"
I think we can all agree that TCP/IP is working out pretty well and that if we made changes to these upper layer protocols, we'd basically be asking the world to undergo a technology refresh because the application layers are adjacent to TCP/IP. So, overhauling layers 3/4 is probably not a good idea, and even if TCP/IP got one, it would not fix the underlying Ethernet problem that created middle hardware switches and routers. Additionally, Ethernet has no capacity for QoS or Broadcast/Multicast. TCP/IP tries to make up for these issues, and again, does a good job...but not good enough to create a fully pervasive system of universal connectivity.
So then, if Dr. Atkinson is saying that we can never get there from here, then what are we to do?
Again, the solution has to provide a migration path that can accommodate legacy machines.
Compatibility is all well and good, but we have to be careful that we don't end up with a major compromise to accommodate it.
What we need is something new that solves a real problem, and then we can worry about a path from here to there. IPv6 has a path, and we're still not moving as fast as we should.
-ASB
I think you are missing an extremely important aspect of what has driven ethernet to survive while others have failed - cost.
As long as a new chipset can be created to support ethernet at faster speeds at a cost that will drive the technology to the doorstep of Corporations - it will remain the predominant player.
Ask the proponents of many other protocols that no longer see the light of day........
That's like saying that, historically, all of my past relationships ended, therefore there's no point in dating. I'd have never gotten married. Yes, it's an uphill battle, but there is always a tipping point and every no gets us one step closer to a yes.
Fortunately most of the majors put their chipsets out in pairs, so the code between the top and bottom is already separated. The trick for us is for all of our drivers to look like Ethernet, and again, for us to be able to run the standard and our protocol at the same time. We'll take a performance hit for running a hybrid network, but our simulations still show an improvement, and more importantly, a potential migration path to a broadcast architecture that can eliminate middle hardware.
I'd have posted a white paper if it was as simple as uploading a PDF, but I have to get back to practicing for my 90 seconds of Silicon Valley fame. Wish me luck.
Events
- Dos and Don'ts of Small Business Marketing May 29 @ 11 am PT
- Lead Nurturing 202: The Next Generation May 31 @ 11 am PT
- The Tricks to Paid Media June 6 @ 11 am PT
- Display Advertising for Brand Awareness June 20 @ 11 am PT
Related Activity
IPv6 solves a good part of the problem, in that the ends will be directly addressable. But it's just an addressing scheme. It doesn't increase utilization of the channel via better sharing, or cooperation between nodes, for say edge load balancing.
The point about interoperability was that Ethernet has been carried forward in the name of interoperability. But then, having rules about never making changes to it also means that there have been no layer 2 innovations in over 25 years! So it took Internet Protocol to become the ubiquitous protocol. But TCP/IP has always needed Ethernet under it in the MAC layer, but really TCP/IP only needs to see the top of layer 2. Now this is a finer point on how the MAC or Data Link layer gets broken down to process the PHY rules, and give access of the data packet to the copper, fiber or air (layer 1, the physical layer)..and even though it's always been flawed, nobody ever went back to fix it. Given that there are a lot of companies making a lot of money on this flaw, you can bet that they not going to make any noise about it.
Back to the point. Bottlenecks are bad and routers create them. If they can be eliminated, we'd all generally agree that would be a good thing. When networks get large and when traffic is high, they generally take a dump because too many nodes are fighting for a channel. This creates a decrease in throughput and an increase in latency which can be responsible for jitter in our conversations and video media. The other problem with contentious based networks like Ethernet is the never-ending incremental overhead for every node added to the network. This sucks up bandwidth in order to keep the network running (overhead), leaving less of the channel for the throughput of your payload data.
Now you may ask, "What's the fundamental problem with Ethernet?" Answer: It is misaligned with our carrier networks which are synchronous. This synchronous/asynchronous misalignment is the fundamental flaw that keeps our data networks from being able to support full broadcast and QoS at the MAC layer.