Colorado Springs, Colo. -- Meru Networks Inc., which has blazed its own trail in 802.11b/g media access control chips to gain system-level advantages, is going for multiple unique attributes in its 802.11n networks.
Most activity in 802.11n concentrates on the capacity and speed improvements allowed by antenna diversity in the access point. But Meru's biggest breakthrough may reside in wireless-LAN control. For the enterprise, WLAN specialists like Aruba Networks and Trapeze Networks, as well as Layer 3 switch players such as Foundry Networks and Extreme Networks, tend to offer either chassis-based or stackable switches.
Meru is going in both directions at once. It is adding its first chassis for WLAN control but allowing blades in the chassis to be used on or off a system backplane. Functions in the control plane and data plane are strictly segmented. The system controller and the datapath distribution point (DP) blade both are based on Intel architectures but can be used in various configurations, depending on the network hierarchy. Meru calls the ability to use controllers and DPs in different tiers "three-tier traffic distribution."
This is not to say that Meru has opted for a vanilla-flavored access point. In the AP, Meru uses its own algorithms to implement channel span, channel bonding and channel layering for optimal 802.l1n capacity and coverage.
All network interfaces comply with 802.11n standards, said Kamal Anand, senior vice president of marketing at Meru, but at the AP and LAN level, the company wanted to provide advantages unobtainable with 802.11n APs based on standard silicon. The three-tier model is meant to take on companies like Aruba and Trapeze with a hardware flexibility that has not been demonstrated in WLANs to date.
On its own, 802.11n is far different from its predecessors. In addition to offering broadband capacity in the hundreds of megabits per second, the new IEEE standard is based on the use of multiple antennas and spatial diversity concepts. With MIMO antenna architectures come certain difficulties, Anand said. Multipath RF patterns and co-channel interference tend to make 802.11n coverage follow a spiked, sunburst pattern, with little of the linear predictability of 802.11a/ b/g. That makes data rates harder to predict in real-world networks, and it makes existing rates fluctuate in unpredictable ways.
Filling the gaps
Meru deploys its APs on a single 40-MHz channel, with channel spanning used at 5 GHz. That fills gaps in coverage, allowing a simple distance rule of thumb (80 feet in an office environment) to be used for AP placement.
Channel layering and bonding buy Meru several capabilities. It can offer a 2.4-GHz 802.11n service by bonding a 40-MHz channel and a 20-MHz channel. Anand expects most enterprise customers will deploy 802.11n at 5 GHz, but he believes the ability to offer higher speeds at 2.4 GHz will help the company serve transitional WLAN networks. Channels can be layered in special locations, such as conference rooms, to achieve small-area capacity of close to 1 Gbit/second.
The system-level effort to create tiered zones of control may be Meru's most radical move. Most chassis-based wiring-closet switches, from vendors such as Foundry Networks or Extreme Networks, are based on Layer 3 Ethernet switches that add WLAN control cards. Meru's MC5000 is a five-slot chassis.
One slot is for a centralized controller for high-level traffic aggregation and processing. The others are taken up by DP cards that perform first-level aggregation from APs. The controller card also can be used on its own in smaller, flatter 802.11n networks, without the chassis. DP cards can be used independently in middle tiers that exist between the controller and the APs.
"The controller performs all traffic management that does not directly impact the inline packets," Anand said. "The distribution points use the datapath hardware that handles the packets at wire speed. We architected chassis and cards so they could be configured independently."
See related image