The 100G controversy

I have worked as a consultant and systems integrator for over a decade. My clients range from "Mom & Pop" retailers to IBM Global Services.

Every time I have delivered a proposal to implement new technology, the litmus test is "Will it place nice with the old systems?"

Since the Gigabit Ethernet connection in my ThinkPad is backward compatible with both 10 and 100 Mbps interfaces, I would EXPECT 100G to play nice with 40G, 10G, and even 1G interfaces.

Let's face it, I may have 1G technology on my computer, but I didn't put a 1G network in the house to connect to my 1.4Mbps internet connection.

For me this just seems like a Rule of Progress. The customer is more likely to adopt new technology that will allow them to ease into upgrades, rather than the typical "Forklift" process.

Comments welcome.

A few facts would have been helpful to the article.

Why the 100G camp thinks they are done.
Why the 100G camp thinks the 40G camp is not done.
Why the 40G camp thinks they are done.
Why the 40G camp thinks that they must be tied to the 100G project.

Would have been informative to see this.

I think the nearly-half speed proposed (40G) could be like a patch. Why not join forces for a much higher speed that can fully address all the current and medium future needs wit a 100G standard? Otherwise, how much time should pass for 40G to become innecesary? Regards

It looks to me like the typical struggle that takes place when advancing any technology. It will take longer to get a working 100G standard than to get a 40G standard because it is a harder problem. The 40G people don't want to wait that long because they see a market need now for which 40G is sufficient.

The risk of creating an intermediate step of standardizing 40G is that the public would jump to it as the next best thing, and then balk at a later 100G technology upgrade because it is too soon to upgrade again for a mere factor of 2.5 gain. 40G would have crushed the chances of 100G succeeding.

But the risk of not creating an intermediate step of standardizing 40G is that 100G might take too long, or cost too much, or be too hard to solve in a reasonable amount of time, thus delaying the solution to an acute business need that could be solved sooner by 40G. If the need is great enough, someone else will come along to fill it with a non-standard solution because they want to make a buck, and that might become an ugly de facto standard without well-designed capabilities.

It is a balancing act of short vs long term needs. The solution requires good judgment about how long it would actually take to finalize each standard. Only someone intimate with the details can really tell that. If it would take only a little bit less time to finalize a 40G standard than to go all the way with 100G, it is probably not a good idea to do it. But if it would take 100G a lot longer to do than a 40G standard, it might be too aggressive, and 40G would be a less risky route. But doing 40G would very likely ruin 100G's chances of success, as markets generally need at least a factor of 10 improvement to want to upgrade in significant quantities.

It is probably a bad idea to try to do both. That would create one standard too many in the long run, and make it harder for future technologies to be backward compatible, and harder for buyers to choose.

If they can work it out so that both standards are on track, one benefit might be that you wouldn't have to aggregate links. Many organizations will use LAN aggregation before upgrading to the next level - for instance, grouping together up to four links of Gigabit Ethernet before upgrading to 10 Gigabit.

If 40G were an option, then instead of aggregating four 10G links, you could more quickly upgrade to 40G, and having a single link is generally regarded as a cleaner approach. And then you could go up to 100G sometime after that as needed.

Jeff Caruso
Network World Site Editor

I'm sorry, but I consider discussion of anything beyond 10G as desperate network product manufacturers desperately trying to create demand. I'm part of a 6000+ employee, 10,000+ device, fully electronic healthcare system and we aren't even stressing 1G links. A lot of people are having a lot of trouble accepting the fact that networking has become commoditized and just isn't ever going to generate the kind of profits it used to.

Its a matter fo the scalability of the PCI bus. 40GB is an acheviable goal on the PC manufacturers side, but a limiting factor from the network infrastructure manufacturer's side.