Fast 802.11n Wi-Fi is going to take off slowly in the enterprise, but all the vendors want to get in the race, as they can see a big increase coming 802.11n does take off. They are getting ready to compete hard - and it starts with the specifications of their new 802.11n equipment.
So who has what?
The first vendor off the starting blocks, Meru, is a special case. Its "channel blanket" architecture puts all the access points on the same channel, and is difficult to compare with the other vendors' approaches.
Apart from Meru, the other significant vendors mostly use an architecture with "thin" access points whose radio channels are planned and managed by a central controller. They are all in the process of introducing 802.11n access points, and upgrading their switches as necessary.
In terms of market share, Cisco is the leader by far with more than 60 percent of the enterprise Wi-Fi market, and is also the first to have its access points certified as meeting the draft 802.11n specification, has the fullest specifications online.
Compared to Cisco, the other players are minnows, with less than ten percent each. The next leading player is Aruba with something like 9.5 percent, followed by Motorola, whose Symbol division launched the Wi-Fi controller sector, established it in its home markets of retail and warehousing, and has been losing market share ever since.
Motorola has announced plans for 802.11n, but not yet delivered, so we won't talk about it further here.
Irritatingly, Trapeze refuses to quote figures to analysts, but is probably still the next largest in share, given its OEM agreements with Nortel (due to end) and 3Com.
Trapeze and Aruba have both announced 802.11n products, and taken potshots at each other, claiming technological leadership over each other and over Cisco. We've spoken to all of them trying to sort out any real points of comparison.
Which silicon is in which AP?
Cisco uses silicon from Marvell, while both Trapeze and Aruba use Atheros silicon in their MP-432 and AP-124/5 access points. Aruba announced its product later than Trapeze, claiming to be using a newer generation of silicon, with better performance, but we doubt this. Products launched within a month or two of each other are almost certainly on the same generation and, as far as we can tell, both companies are using the Atheros 9160 Wi-Fi silicon. (Meru, we're told uses Atheros also but, with products out before the others, it is very likely an older version).
In one or two measurable ways then, Trapeze and Aruba's APs seem likely to outperform Cisco's. The Cisco 1250 AP will handle 2x3 MIMO (two antennas at the client and three at the AP), while both Trapeze and Aruba can handle 3x3, on a single radio or on two.
Cisco’s 11n Q&A section plays this difference down, saying "Theoretical calculations show that a 3x3 radio will have slightly better performance over a 2x3 radio. However, in our real world performance testing we have not seen any significant performance differences between 2x3 and 3x3 implementations," pointing out that both configurations support the multiple data streams that give MIMO systems better throughput.
Michael Coci, directof of technical marketing at Trapeze, denies this, saying that there are differences, which can be summed up as "a 3x3 will give a higher data rate at longer distances than a 2x3 implementation".
But that's not the end of it. The performance of an 802.11n access point depends critically on the ability to power it - and enterprise access points are usually powered over Ethernet, to avoid the labour and expense of running extra power lines into the ceiling cavities where they are installed. We've gone on about powering 802.11n at some length (and here), but it's a crucial issue and more facts are emerging.
Currently, all dual radio access points need more power than can be delivered by the letter of the 802.3af power over Ethernet specification. The 802.3af specification offers 15.4W of power - but may be as low as 12.95W at the end of a cable run; Cisco's 1250 AP requires 12.95W when it has one radio in operation, or 16.9 W when two are in use. Aruba and Trapeze don't publish such dctailed sheets, but they seem to be in the same league, both quoting 17W for a dual-radio AP. If you want to run two radios, it seems you need more power from a local power brick.
It's a problem which will go away in the next year or so. (and this may be the biggest reason why people will hold off on 802.11n). Successive revisions of the silicon may use less power, by integrating the chips more tightly - but more importantly, the 802.3at Power over Ethernet Plus specification, due to be competed in 2008, will deliver more power over a cable. Early drafts suggested around 60W, but more recent sources seem to promise around 30W - but either way, it's plenty to power a dual radio AP.
Till that arrives, however, power is a problem for 802.11n APs, and all vendors are in the same boat. But they all have radically different responses.
Cisco's answer, typically, is a clear, definitive answer - and one tha tinvolves more networking equipment. It has a proprietary power injector that goes beyond the 802.3af specification and can power a dual-radio AP. It will replace this with an 802.3at injector when the standard is ratified.
"While Cisco would have preferred to use 802.3at (PoE+), the standard is not expected to be ratified until end '08/early '09, so we needed to design our own interim solution," explained Pat Calhoun, CTO of Cisco's wireless networking business unit. "Through some Cisco innovation, we have designed a system that allows our APs to provide full service even when powered over Ethernet. Unfortunately, this is not a standard interface and we recognize this is not optimal, which is why we are working the standards angle hard to accelerate the process." Letting a customers make investment decisions only to find out later on that their network is operating at a lower performance is, he says: "less than ideal."
Trapeze came up with a clever solution. Its AP has the ability to take power from two separate Ethernet ports (Trapeze and Aruba's enterprise APs both have two Ethernet connections for redundancy). This may mean using more wires and connecting more powered ports required at the wiring closet, but eliminates the troublesome local power brick.
Aruba, when it launched its products, muddied the water by promising two apparently distinctive things. Its AP can fall back to a 2x3 configuration, on two radios, when only 802.3af power is available. And - the company claims - the 17W power requirement of a full dual radio implementation may be beyond the official limits of an 802.3af power implementation, but in practice, most of them can stretch that far.
Trapeze has poured scorn on both of these. As the companies use the same silicon, these options are available on Trapeze access points, says Coci, "the 2x3 option is possible on the MP-432, but we did not mention it in the product materials because we didn’t think it was a reliable solution." Similarly, the MP-432 uses the same power level, but Trapeze doesn't recommend pushing PoE systems to the limits.
What Aruba's 802.11n AP can't do is draw power from both Ethernet ports. The company saw this as an interim solution and didn’t add it.
The companies' strategies on encryption is also a big differentiator. All WLAN traffic has to be encrypted when it goes over the radio waves, but Aruba encrypts its traffic at the controller, while Cisco and Trapeze encrypt it at the access point. This means that all Aruba WLAN traffic has to pass through the central controller, while Trapeze has been able to distribute some authority to the access point, and not all WLAN traffic has to go through a controller.
The benefits of these approaches are hotly argued, with typical marketing exchanges. Trapeze only distributes traffic, says Aruba, because it can't make a big enough switch to handle it centrally. Aruba only handles security centrally to boost the perceived value of its switch.
This sort of argument is entertaining - for a short while.
"The radio chip is only one element in the magic of our AP," says Aruba's head of technical marketing, Michael Tennefoss. "Power management, encrypted packet processing, security certificate validation, redundant GE ports, downloadable configuration, and many other important characteristics distinguish our APs from those made by Trapeze."
Aruba has built a Cavium network processor into its APs, which it says gives them greater abilities than rival products.
Trapeze claims to have built its own antenna system, with advantages over the standard antennas that Aruba uses. If nothing else, Trapeze's antennas are hidden neatly inside the unit, which looks like a smoke detector.
Trapeze is also proud of its Ringmaster planning system, which it has upgraded to a special 802.11n edition, which takes account of the increases in range and penetration produced by the new wireless standard.
Both companies are unimpressed with Cisco's unit, of course.
"The Cisco 1250 is not richly featured," says Tennefoss. "It’s heavy (2.3kg), physically large, requires special high power PoE, does not offer hardware assisted encryption, cannot be repurposed by download, only has a single GE port, and doesn’t offer client-to-core security."
All these seem to be accurate according to the Cisco data-sheet.
A year of marketing
The differences between the equipment may seem small, but vendors will do their best to maximise the benefits they offer over the next year, with very few actual case studies to illustrate those benefits.
It's going to be a grim year in enterprise Wi-Fi marketing.