I’ve come to the conclusion that the adoption of 100 Gig to be much bigger than 40 Gig. To me, 40 Gig was really a step on the way to 100 Gig. Ethernet has historically jumped in logarithmic steps and a 4x jump just doesn’t seem like the bang is there for the buck spent on new hardware and upgrades.
The standard for 100 Gig-E was ratified last year and since then we’ve had a number of vendors, the usual suspects – Juniper, Cisco, ALU and Brocade, launch 100 Gig-E line cards. Recently, I had a chance to discuss the topic with Greg Hankins, Global Solutions Architect for Service Providers at Brocade. I thought I would share some of my thoughts on the 100 Gig-E market.
First, the when. Clearly we’re in the very early part of the market. To use a baseball analogy, I think we’re not even out of the first inning, yet. Late 2009 and 2010 were big years for testing 100 Gig so the game is definitely afoot, but only the early adopters are playing. This means companies that need a lot of bandwidth such as service providers, web companies, film and media, high performance compute, financial services and research environment. Adoption will be limited to these verticals for the next five years or so as many general enterprises are in the midst of deploying 10 Gig. The one exception might be as backbone connection for some larger companies but the deployments will be small.
The cost of 100 Gig is likely to remain high for the short term as well as we’re not likely to see the merchant silicon vendors have 100 Gig chips for a few years. 100 Gig is very processor intensive as it requires a packet lookup every 6.7 nanoseconds or 10x the 67 nanoseconds that 10 Gig required. This will limit the solutions to the mainstream network vendors that have the capacity to build their own ASICs. Brocade has the low cost solution at about $100,000 per port but can be as high as several hundred thousand from Cisco and Juniper. The optics are an additional $70,000 to $150,000 for the optics depending on the type of optics required. This is where there is room for some short term cost reduction. The current optics can provide connectivity at 150 meters, 2 kilometers, 10 kilometers and 40 kilometers.
Now, for how. The most cost effective way to deploy 10 Gig is to try and do the upgrade in conjunction with another network initiative. Some examples are deploying 100 Gig along with IPv6 while going through consolidation of data centers or when rolling out a new service. This doesn’t cut the cost but the tie in to another initiative can help justify the investment.
Additionally, I think it’s wise to only upgrade what’s required for the next three years or so as 100 Gig is still in it’s infancy. The next generation of 100 Gig will bring greater density to the 100 Gig line card optics. The current CFP (C Form-factor pluggable transceivers are about the size of an iPhone 4 limiting the density to about 2 per line card (See below). The next generation of transceivers (CFP2s) will be about half the size enabling a card with 4 ports on it.
The industry is in the very early stages of this so we’re not likely to see the second version of 100 Gig for a few years from now.
As a general rule of thumb, to know if you’re 100 Gig ready or not is to consider the number of 10 Gig ports that are being for connectivity. If you’re considering deploying 6-8 10 Gig-E ports, then it’s time to consider 100 Gig. This may be a bit more expensive but the complexity of managing 6 connections is considerably more difficult than a single connection.
So, clearly 100 Gig-E isn’t for everyone today but it’s certainly applicable for about a quarter of the market today but it’s good for most network managers to at least start the educational process today.