In our belief, the immediate future of networking involves the commoditization of 10 Gigabit Ethernet (10GigE or IEEE Standard 802.3ae-2002) and related physical and logical technologies.
The development of 1 Gigabit Ethernet (GigE) over the last decade has shown us how the industry will tumble into its 10 Gigabit replacement. Switch manufacturers, always the first to the battle lines, came out a few years ago with products that sported a very close variant to the 802.3ae standard. A simple firmware upgrade is all that will be needed, they always say, when the standard reaches final status. And for the most part, they were right. OEM manufacturers were next, often the same companies as the switch manufacturers. As soon as the chipset to packet and depacket 10GigE arrived on the scene, clone OEM manufacturers started copying and cheapening the chipsets even further. The Intel Reference Motherboard, often referred to as the Golden Chalice of standards compliance, still does not contain 10GigE networking, and won’t for some time until the prices to manufacture the related chipsets come down.
When 10GigE comes of age, it will allow us to recreate networking infrastructures that currently run at GigE. Some will say that the new cabling standards of CAT6a or 7 will be needed to efficiently run 10GigE, but others have touted that CAT6 cable will be able to run 10GigE for most applications. Facilities that are running their cables longer than 50m (the standard has a maximum of 100m), might face challenges or have to rewire.
So why all the geekdom about this standard? It’s just faster networking, right? Take note of the speed of 10GigE: it leapfrogs, at least in theory, the entire speed evolution of Fibre Channel. FC is currently at its 8Gb/s version, with only one further standard evolution on the horizon: 16Gb/s. 10GigE has 40 and 100 gigabit variants already on the table.
But let’s not take away the significant benefits that Fibre Channel brings to the world of storage networking, specifically two: guaranteed packet delivery and guaranteed packet order. Even though Ethernet is about to usurp Fibre Channel in speed, it doesn’t have these guarantees as part of its spec.
Enter a new standard: Fibre Channel Over Ethernet, or FCoE. INCITS, or the International Committee for Information Technology Standards (pronounced insights), developed this one, and FCoE is in its final status before ratification. FCoE will allow true Fibre Channel communication across standard high-speed Ethernet networks (10GigE or higher), and contains a provision for FCoE capable devices to discover each other over an Ethernet network. FCoE doesn’t require anything other than hardware or software-based initiators to be developed for the computers and storage that will be in the network, and they can all use the same Ethernet network that the rest of the facility uses. FCoE will not be able to route, however, so all the machines in the FCoE system will have to be within the same subnet.
We’ve already mentioned we have the tools to bring standard networking to 10GigE. We’ll still have to wait a while for the commoditization of the standard, and with it, affordable pricing. Rewiring will also be daunting for many facilities, but one consolation is the concept that 10GigE also can run on glass. This means that runs of optical cable, formerly intended for Fibre Channel networks, can be reused to talk 10GigE. But the next wave of adoption will truly come when storage manufacturers, always last to the dance of a new networking standard, offer up storage controllers that can talk FCoE at 10GigE.
Which brings us to the future of SAN topologies in the next three to five years. We see various disparate systems converging into 10GigE networks, with protocols floating on top in order to provide the critical services necessary to deliver the data in the way that people and processes will need it. While the availability of storage devices that can talk FCoE will be the true catalyst of this change, smart IT administrators will begin early. They will rewire and buy “FCoE-ready” switches within the next few years, so that when the storage comes to market, they will rather seamlessly transition into this new networking ecosystem.
Wow, eight paragraphs of networking gobbledygook and no talk of Macs. Ok, here’s the Mac paragraph. The new Mac will have one of two options for networking. Apple might be a trailblazer here and offer a copper or optical 10GigE port on the back of a Mac, but only when the standard becomes commoditized to the point where that hardware will be cheap to integrate. More likely, they will wait for a third party to offer a 10GigE adaptor that will plug into the Light Peak port (more on Light Peak in our previous article in this series).
So, for those wondering where the Fiber Channel or 10GigE card will go in a Mac that has no PCIe slots, hopefully this assuages your concerns. We won’t need the card, as long as we have a native 10 gigabit peripheral bus converting to a 10 gigabit networking standard. And perhaps, just as Light Peak reaches its 100 gigabit variant, so too will 100GigE be sweeping the industry as the next networking plateau.
The software tools that creatives will use in the future? In our next article...
Comments