Aerohive Networks removes the central controller, putting the control plan along with the data plane in the access points (APs) themselves, thereby obviating the need to backhaul traffic to a controller for forwarding, said CEO Dave Flynn.

Revenge of the fat AP?

Of course, the first generation of WLAN APs, the so-called fat generation, also had all the intelligence required for them to work on board, and plugged directly into a corporate LAN. The security issues raised by this approach were what spawned the next generation of switched WLAN networks, in which thin APs backhaul to a dedicated WLAN switch on an overlay network that is thus isolated from the main LAN.

That generation, whose primarily proponents were Aerospace (now Cisco), Aruba and Trapeze, always faced the challenge of scalability, however, so in recent years we have seen the evolution of controller-based architectures with traffic backhauled over the corporate LAN to a dedicated controller, which may be a standalone device or a blade on a wired switch.

Fast roaming and best-path routing

As such, Aerohive’s architecture may sound, a priori, like a throwback to the days of fat APs, with the difference that far more intelligence is now being placed in Santa Clara, California-based Aerohive’s APs than in its fat forbears. For instance, Aerohive has a feature called Fast Roaming, defined as sub-50ms handoff between APs. Another is Best-Path Routing, whereby the APs have the smarts in them to choose the best route through the network for data coming from wireless endpoints.

Apart from the efficiencies inherent in endowing what amounts to the network on-ramp, i.e. the AP, with the ability to work out which path data should take through the network, this approach brings another notable benefit. Since the APs autodetect each other, one can go down with the others failing over in sub-50ms times, enabling Aerohive to claim high resiliency for its architecture.

Mesh capabilities

Furthermore, since its APs are all dual b/g and a devices, they can be deployed in mesh mode, in which case a proprietary fast routing protocol kicks in to provide similar fast failover functionality for the mesh nodes.

Flynn was careful to point out, however, that Aerohive is not entering the market to compete with dedicated mesh vendors like Tropos, Belair and Firetide, however. They are going after the major outdoor deployments such as in the municipal WiFi networks that have already swept the US and are now catching on in Europe, whereas Aerohive’s meshing capabilities are more limited, the idea being for a couple of APs to be deployed in mesh mode to extend the network, for instance, into a particular area of a building where coverage is desirable but pulling cable is not economically viable.

L2 and L3 roaming

On the question of roaming, Flynn said there are two different scenarios that any WLAN infrastructure needs to address. The simpler one is a Layer-2 roam between one AP and another on the same subnet. In that case, Aerohive does what it calls predictive roaming, i.e. it primes the next AP to receive traffic from a device in mid-session, before it actually arrives.

In the case of Layer-3 roam, i.e. from one subnet to another, we do what we call Dynamic Network Extension, of DNX, which means we establish a generic routing encapsulation [GRE] tunnel back to the original subnet so the device doesn’t need to get a new IP address in mid-session, he explained. We remain at Layer 3 only as long as necessary, in other words till the end of the session [i.e. once a voice call is completed], then we change IP address and Best-Path Forwarding can resume.

Pricing

The Aerohive portfolio consists of two products. The first is the HiveAP 20, described as a dual-radio AP with fast stateful roaming capabilities that, at $995 list in the US, is clearly more expensive than APs from the likes of Cisco, which range from $699 to $899, said Flyn. However, once you factor in the additional cost of the controllers, we work out cheaper, he went on.

Second is the HiveManager, which is not a controller, since the control plan resides on the APs themselves, but is used for configuration and user policy enforcement, in other words, it is houses the management plane. The US list price there is $4,995.

The marketing pitch

Aerohive’s marketing focuses on the fact that controller-based architectures entail all the data being backhauled through the network to the controller, which raises the potential for latency (a major problem for voice traffic) and only tends to get worse once 11n, with its two- or threefold increase in wireless bandwidth, is standardized and enters the enterprise market.

As a tactical marketing approach, meanwhile, the company is highlighting the fact that, as Flynn put it, we are a plug ‘n’ play upgrade to fat APs, in that its devices can be dropped in as swap-out replacements, bringing not only WPA encryption rather the weaker WEP that characterizes most older APs, but also enhanced manageability with the HiveManager and all the mobility features, QoS and queuing functions Aerohive has built into its access points.

Aerohive vs. Cisco

No surprises, then, that it is targeting the huge user base of Cisco’s AiroNet fat APs as its low-hanging fruit, with Flynn also arguing that the networking giant has effectively ceased development on the old platform in favor of the Aerospace thin-AP architecture it acquired a couple of years back.

To accompany its launch, therefore, Aerohive has prepared some cost scenarios against Cisco Aerospace kit. Using an unnamed customer with a corporate HQ and 800 branch sites, it calculates initially the cost for what it calls the convenience scenario at HQ, in which redundancy is not a factor, concluding that 200 Cisco APs plus two of its AC100 controllers would cost $299,000 compared to 200 HiveAPs at $199,000, but that the difference increases with redundancy, because our APs have redundancy built in.

In other words, if an additional Cisco controller is added for the purpose, that puts another $35,000 on top, while if high capacity requirements are also to be met, it adds 18 smaller controllers (the AC25) and another 200 APs to bring the grand total to $644,000. This compared to a cost, with an additional 200 HiveAPs, of $398,000.

For the branch sites, the difference are $3.4m for Aerohive and $5m for Cisco in the convenience scenario, and additional $2.6m for Cisco with redundancy and a total of $12.9m for Cisco and $6.8m for Aerohive with high capacity, with the Cisco calculations all based on the smallest of its controller range.

Our View

Aerohive in entering a market dominated by Cisco as the incumbent in half or more of all accounts, but with a number of challengers including the recently IPO’d Aruba, Trapeze, Symbol (now Motorola) and Colubris, not to mention the coverage blanket proponents Meru and Extricom. As a start-up, it needs to make a lot of noise to get noticed, and have some pretty compelling arguments as to why enterprises should swap out their tried and tested suppliers for some new kid on the block.

It counts, for this purpose, on a couple of perceived inflection points in the market. Firstly, more and more voice is or soon will be going over WiFi infrastructure, with far less tolerance for latency than data traffic. Second, 11n is upon us, with the expectation that it will start to make its way into enterprise networks in 2008, and thirdly, there is the fact that WLAN is moving from being a nice add-on for convenience and guest access to become an integral part of companies’ infrastructure, whether in conference rooms, out-of-the-way corners of a campus, or simply in the cafeteria.

In this context, targeting the Aironet replacement business clearly has its merits. That a new start-up will have its work cut out getting onto major customers’ radar, however, suggests Aerohive will need some big backers such as enterprise SIs. If its architecture is truly as successful as it claims in delivering functionality for the next wave of 802.11 networking, it could well become an acquisition target for anyone that wants technology to challenge Cisco in enterprise WLAN, or even for Cisco itself…