chris - Fotolia
There's one thing Joe Rogers can count on every January. Actually, it's more like thousands of things. All of the new gadgets that students at the University of South Florida receive as gifts over the holidays are guaranteed to pop up on his wireless network when the new semester starts.
Students return to the Tampa campus showing off shiny new laptops, smartphones, tablets, wearable devices, smart TVs, Blu-Ray players and gaming consoles, nearly all of which are trying to connect to the university's 4,000 wireless access points (APs). An increasing portion of those devices, about 10%, are compatible with the latest Wi-Fi standard, 802.11ac Wave 1, which provides a maximum theoretical speed of 1.3 Gbps in the 5 GHz spectrum.
Rogers, the university's associate director of network engineering, has deployed about 1,600 APs that support 802.11ac Wave 1. And, as in most enterprise networks, the university's wireless traffic is backhauled from the APs over Cat5e and Cat6 cabling that connects to1 Gigabit Ethernet (GbE) edge switches in the campus LAN. It's a suitable arrangement for the time being, since those links are nowhere near saturated, partly due to the small number of 11ac-compatible consumer devices on the market.
Of course, that will change. The second and final version of the standard, 802.11ac Wave 2, ratified by the IEEE last year, is designed to reach speeds up to 6.93 Gbps. Although no natively Wave 2-capable APs are commercially available yet, Rogers, like many network engineers, is already thinking about how or even if they might work in his network. A single gigabit backhaul link couldn't take full advantage of Wave 2's maximum throughput. Then again, deploying 10 GbE in the access layer wouldn't make sense for several reasons: The interfaces are more expensive; the current 10GBASE-T specification doesn't include Power over Ethernet; and it could even require ripping out legacy cabling and replacing it with Cat6a (that little "a" stands for augmented performance), even though that's unlikely.
Joe RogersUniversity of South Florida
"Even if 11ac Wave 2 access points had a 10-gig NIC, what switch am I going to buy?" Rogers asks. "Not to mention, can I afford it?"
That's why recent news about efforts to develop specifications for 2.5 and 5 GbE over legacy cabling caught Rogers' attention. While the prospect of having two alternatives between 1 and 10 GbE is compelling on its own for many enterprises, at the heart of what's driving interest in the two new Ethernet speeds is the intention to support them over Cat5e and Cat6 -- two of the most common types of cabling found in campus LANs today.
Last fall, Cisco and Aquantia announced they were developing specifications for 2.5 and 5 GbE and forming a vendor consortium called the NBASE-T Alliance, its representatives contending that standards bodies are moving too slowly as 802.11ac Wave 2 lurks around the corner. Just a few months earlier, however, Broadcom and other vendors including Avaya, Aruba Networks and Brocade had announced their own consortium, the MGBASE-T Alliance, which has the same mission and is writing its own specs.
Caught in the middle are the IEEE, which recently launched its own study group to work on creating new universal standards for 2.5 and 5 GbE, and the Ethernet Alliance, a nonprofit that advocates for Ethernet standards development and interoperability. Each vendor consortium has reportedly made efforts to coordinate with the IEEE and the Ethernet Alliance behind the scenes, but due to the formalities in the standards development process, vendor camps will likely outpace the progress of IEEE. Both vendor groups are rumored to already have designs ready for pre-standard products.
It marks the second time in a year that vendors have initiated the development of new Ethernet speeds, conjuring up a sense of déjà vu after a consortium led by Arista, Broadcom, Google, Mellanox Technologies and Microsoft announced last July they were developing specifications for 25 and 50 GbE independent of standards bodies. The move came after the IEEE had rescinded a call for interest for the two technologies, citing low attendance and lack of a quorum at a meeting last March. Members of the vendor consortium said cloud providers had an urgent need for 25 GbE in their data centers and couldn't afford to wait for the IEEE. The standards body later voted to form a study group.
"We're right back to the old days of one vendor with a lot of partners seeing if they can drive through a standard by sheer force of will, or just shipping products before a standard is developed and hoping that it becomes a standard after the fact," says Loring Wirbel, a senior analyst at the Linley Group, a semiconductor-focused market analysis and consulting firm in Mountain View, Calif.
That's exactly what worries network engineers like Rogers, who has a multi-vendor network.
"My real concern is the competing alliances right now. As a customer, I want to see one solution that works across all of my vendors," Rogers says. "I'm happy to see them doing it, and I may have areas where 2.5 or 5 is useful to me. But it's not going to be useful if there are two different standards -- if my Brocade switches won't talk to my Cisco access points."
It's an unlikely scenario for enterprises, at least in the long term, according to Wirbel, who says vendors in these situations typically guarantee to customers that future products will be compliant with industry standards once they emerge. Additionally, since the initial consortium announcements, some vendors, including Aruba and Brocade, have pledged their allegiance to both sides.
"Nobody has ever been left high and dry," Wirbel explains. "The real question to ask is, does the creation of two alliances like this throw all kinds of monkey wrenches into the works and make it a longer process for standards development? That can happen. It's just inevitable. You've got two big camps here with a lot of egos involved."
Squeezing more out of Cat5e and Cat6
The sheer increase in throughput between 802.11ac Wave 1 and Wave 2, while significant, isn't the only thing driving interest in new Ethernet standards. The Wave 2 standard introduces a new capability, multiuser MIMO, which would enable compatible APs to transmit multiple streams of data to several clients simultaneously, whereas current standards only allow single-client access.
"You need more than one gig but less than 10 gig, and that's where the need came for something in between the two," says Simon Assouad, president of the MGBASE-T Alliance. "It's primarily driven by multiuser MIMO technology, and this is really to make sure an access point doesn't have the wired connection be the bottleneck to the switch."
John D'AmbrosiaThe Ethernet Alliance
Under today's specifications, Cat5e and Cat6 cable are certified to support Gigabit Ethernet at up to 100 meters, and at up to 100 MHz and 250 MHz, respectively. That's fine for 802.11ac Wave 1, but if enterprises were to deploy Wave 2 APs with that infrastructure today, they would undoubtedly run into performance issues, according to Drew Lentz, a wireless network solutions engineer at Frontera Consulting, a systems integrator based in McAllen, Texas.
"That bottleneck isn't just about the switch that it's connected to. It's about the limitations of the cabling that are connected to that interface," Lentz says. "Using the proper cabling and proper connectors to transport the data is absolutely imperative because you want to make sure you can take full advantage of the 802.11ac offering without having something like the cabling or connectors slowing you down."
Although Cat6 cabling can support 10 GbE, it can only run at limited distances. Supporting higher speeds in the access layer would require Cat6a, which is certified to support 10 GbE at up to 100 meters, but that kind of rip-and-replace is cost prohibitive for most enterprises.
"If you're trying to retrofit an old building, it gets pretty complicated to go change the cabling," says David Chalupsky, chairman of the Ethernet Alliance's BASE-T Subcommittee and acting chairman of the IEEE's 2.5/5 GbE study group. "But the market is finding that, yes, we can really get more than 1gigabit of bandwidth out of the installed base of Cat5e and Cat6 cabling. From a technical standpoint, it's very feasible."
And while many IT departments don't have a need for 802.11ac Wave 2 today, it's only a matter of time before that day comes. Having a wired network that's ready for it will make the transition much easier, Lentz says.
"We still don't have as many clients using 11ac as those using 11n, so it's not really taxing the network as much as it could be. But that doesn't mean there's no deluge of data that's going to hit us in the next three to six months," he says. "When that happens, that's when people are going to start to find all these issues in their backhaul links. It's like driving down a street in a city that was built many years ago, and they never anticipated all the traffic that would be on there after they opened up a Wal-Mart."
Two approaches: What's the difference?
Both vendor consortiums are working toward the same end result -- running 2.5 and 5 GbE over existing Cat5e and Cat6 cabling. But the difference lies in what aspect of the technology they tinker with to get there. Assouad, of the Broadcom-led MGBASE-T Alliance, declined to go into detail about the consortium's strategy, saying only that it was "based on improvements to the 10GBASE-T specification." There are indications that Broadcom is focusing on clock speeds in chip designs to slow down data rates in 10 GbE, but nothing has been publicly confirmed, according to Wirbel.
Spokespeople from the Cisco- and Aquantia-led NBASE-T Alliance say they initially looked at slowing down 10 GbE, but it was too problematic. Cabling in many buildings is often 15 to 20 years old, so the specification needs to account for possible degradation in the cables and the presence of lower-quality connectors, according to Kamal Dalmia, vice president sales and marketing at Aquantia.
"10GBASE-T is based on signaling called DSQ-128, which is based on PAM-16 technology. It has this notion of uncoded bits that are susceptible to reflections that come from discontinuities in the channel, and an example of a discontinuity would be a connector, so that's why keeping that DSQ-128 structure would not be a good idea because of the older, poorer connectors," Dalmia explains. "What we did was retain the PAM-16 but ensured that there are no uncoded bits in the system. So we have what's called a fully protected system, as opposed to a partially protected system."
But does the simple fact that vendors are divided mean enterprise network engineers will be at their mercy, struggling with interoperability between proprietary interfaces in multi-vendor networks? It's extremely unlikely, according to experts who have lived through the development of Ethernet standards before.
"The reality is that when you go into any of these projects, there are always competing proposals. So we have two competing proposals -- OK, big deal. We've had that before. It's just that now you have alliances there," says John D'Ambrosia, who, along with being the founder and chairman of the Ethernet Alliance, is also chairman of the IEEE's 400 GbE task force. He previously chaired the task forces that developed the specifications for the 40 and 100 GbE standards.
"We're trying to facilitate and reach out to these organizations to express our concern that, ‘Look, we need you to understand that our members are very concerned, that they want to see one standard and that there be multi-vendor interoperability,'" D'Ambrosia adds. "We're trying to get them to understand that, ultimately, the success isn't in ‘winning' the standard. The success of anything is really when it gets deployed."
Optimizing 10 Gigabit Ethernet: Why it takes more than hardware
Getting wired networks ready to support 802.11ac
What the ratification of 802.11ac means for you