Cisco recently introduced another addition to its Radio Resource Management (RRM) algorithm – Dynamic Bandwidth Selection (DBS). Previously, controllers only had the ability to automatically assign channels and power levels, while the AP channel widths were manually assigned for each AP Group. A choice of 20MHz, 40MHz, or 80MHz channel widths were offered within the Access Point (AP) groups, allowing for the APs to all be on the same chosen channel width. With the addition of DBS, the APs can now be automatically assigned to individual channel widths. This allows the radio resources more opportunities to fine-tune the network according to changing RF conditions for higher granularity.
Independent testing company Miercom recently carried out a full scale RRM testing to run Cisco’s DBS RRM against Aruba’s Adaptive Radio Management (ARM). This testing was performed to validate and benchmark the best-in-class radio resource management feature in Enterprise Wireless. The RRM testbed was designed to achieve real world apples-to-apples comparison between the Cisco 5520 controller with five AP 2702i against the Aruba 7210 controller with five AP 225.
The result was a runaway win for the Cisco controllers over Aruba.
RRM is a vital element for strong performance and reliability of a wireless network. RRM involves intelligent management of the channels and power transmission of AP radios based on the environment Radio Frequency (RF) conditions. Cisco introduced RRM on its WLAN Controllers in 2005, and since then has been an industry pioneer in automatic AP channel and power management. Cisco’s RRM has evolved over the last decade to become one of the most important WLAN features. RRM is especially helpful for high-density networks where scores of APs need to assign the best possible channel and power plans. These channels and power plans create minimum RF interference and maximum RF performance.
Miercom independently verified the functionality of both the RRM and ARM over a period of several days to analyze the overall performance impact on the network. They observed each controller with different channel width conditions ranging from 20MHz to 80MHz. Typically the radio management algorithms for both RRM and ARM take a few hours to zero-in on what is considered to be the best channel and power plan by the respective vendor controller. Miercom allowed each test setup a period of 12 hours to settle down to the best possible AP channel and power plan.
One of the most interesting observations was that the Aruba controllers would require a complete reboot to correctly trigger the ARM algorithm. This meant that the Aruba APs would fail to choose a reasonable set of channels and power levels if the controller was not rebooted. Without the reboot, the Aruba controller would assign the same channels across multiple APs even after completing the full settling period. Although rebooting the Aruba controller and allowing another attempt to settle helped the APs to choose a much wiser set of channels, the Aruba product still ended up with substantial overlapping of APs on the same channel. On the other hand, Cisco infrastructure was pretty quick to adapt to the different channel width conditions and swiftly kicked into the RRM algorithm to determine a sensible channel and power plan without the need for any sort of rebooting.
With the UNII-2 and UNII-2E (DFS) channels disabled to create a fair chance of overlapping among the five APs on both vendors, it was interesting to see the difference between the final channel-planning for each vendor. Cisco’s channel planning always resulted in the least possible overlapping, while Aruba consistently suffered from heavy overlapping of APs on the same channels. This is evidenced in the Channel selection visual representation charts below showing the 40-MHz channel width tests.
The channel selection visual representation chart below indicates how Cisco’s Dynamic Bandwidth Selection for Auto channel widths yielded the optimum channel and channel-width planning. Cisco came through with 0% overlapping among any of the five APs. DBS intelligently chose a combination of 40MHz and 20MHz channels which resulted in an excellent frequency reuse with each AP having an individual channel to operate on. Since Aruba did not offer any auto channel width configuration, there couldn’t be a direct comparison to the Cisco controller.
As exhibited in the performance graph below, overall performance for each test case stood out to show the client performance densely suffered on the Aruba infrastructure due to ARM’s poor channel selections which catered to higher retry rates, co-channel interference, and reduced channel capacity for every test case. Clients on Cisco were able to achieve a considerably superior throughput performance across all channel widths due to the excellent reuse of the channel frequencies, thriving in a performance friendly environment. Not including the Auto Channel Width Selection Test, Cisco trumped Aruba by an average of 200.3 Mbps in throughput on every test—including a robust 341 Mbps in the 40 MHz channel width testing.
For more details on the test setup, clients used, and Cisco RRM with DBS, check out the complete Miercom Performance Report on Cisco’s new 5520 and 8540 wireless controllers here.
In case you missed out, here are the previous blog entries detailing the insights for the Data Plane Tests and Controller wired throughput comparison tests validated by Miercom.
Power management still depends on knowing your clients Tx power since no standard excists for aligning this in regards to two-way wireless communication like colaboration. Having a single AP transmit at highest regulatory power level might ruin transmissions. DTPC is proprietary and therefor only works for CCX capable clients. Maybe an algorithme looking in to the retry rates seen from clients and APs could solve that within an AP group. Performing a manual RF channel plan still optimizes airtime better in high density environments. But one could use wlc channel planning as a start since optimizing for 802.11ac require more cell tuning for the high MCS rates.