Unleashed experiences from a ZD user

[john_d]

Up to this point, I've been a happy ZoneDirector user for almost 3 years. I have been looking at Unleashed very closely since its announcement, since my two deployments are both home / home office, and the costs for ZD + support + APs is relatively steep to begin with, and even steeper if you find a single point of failure (e.g. controller failing) unacceptable.


At any rate, this weekend I finally got a chance to test out Unleashed 200.2 GA in a larger scale deployment. This was my family's house. They've got a 6000 sq ft house with about an acre of land. Furthermore, the architecture of the house makes it cost prohibitive to run ethernet beyond the basement, making SmartMesh an attractive choice. They are currently using a ZD1105 + 5 ZF7982's (Two root APs in basement, 1+1 MAP-eMAP pair on first floor across a long bookshelf, 1 MAP on second floor). The eMAP configuration helped immensely with the bandwidth at the 2nd hop, which is no surprise. I cannot say enough positive things about Ruckus's eMesh hybrid arrangement for high-throughput meshing. 

When I first set this up 3 years ago, their internet was 20mbit, and the network delivered 35+mbit everywhere, even to 1-stream handheld clients, so it was good enough. Since then, they've upgraded to 100mbit and it bothered me that their wifi was a bottleneck for the network. Since I upgraded my personal configuration to R710's, I had some spare Unleashed-capable AP's to gift. 


Long story short, I was able to replace their 5x7982 setup with a R600, R600, R500 setup, and it delivers about double the speed of the original network. I have some expansion plans for my next trip, but for now everyone is a happy camper. As far as Unleashed goes, here's some of my thoughts as a ZD user trying Unleashed for the first time:


The Good:

• Ruckus really did a good job with the Unleashed UI. Compared to ZD, the initial setup was markedly more straightforward and it might even be possible for non-IT-professionals to set up Unleashed by themselves. The wizards actually guide you to a useful default configuration, while with ZD you have to drive some configuration yourself after an install.
• Performance felt pretty indistinguishable to ZD, for moderate density. They have around 30-35 wifi connected devices, including bandwidth-hungry concurrent 4K streaming boxes. The R600/R500's performed quite similarly to when I used them under ZD1200+9.12, and I did not observe any capacity issues. Furthermore, compared to a ZD1100, the WebUI responsiveness is way up. I was wary that the Unleashed Master takes on ZD functions which reduces its capacity by about half, but it did not translate to noticeable impacts for my use case.
• Overall, I think the cost of Unleashed makes a lot more sense to SMB. The ability to expand past 5/6 APs without additional licensing fees is great for SMB/SOHO, where one generally likes to start frugally and expand as time goes on.

Suggestions for the future:

• Please consider adding some form of SpeedFlex, at least between AP's. Compared to ZD, mesh AP placement tuning was a lot more cumbersome with Unleashed due to the lack of SpeedFlex for measuring mesh link throughput in realtime. In the end I was able to use iperf / SpeedTest on a client, but it took a lot more effort to ensure I was connected to the desired AP, the desired band, and physically move to the AP. In a 6000 sq ft house, I literally walked 2 miles in the process of setting up the network!
• I missed the ZD radio stats (especially airtime percentage, percentage non-unicast). It can still be accessed via the supportinfo, but it takes a lot more digging.
• Having Refresh buttons for various parts of the UI would be nice. Unleashed auto-updates at ~1min intervals, but sometimes it helps to monitor in realtime and hitting Refresh in the browser causes you to lose your spot.
• It'd be great to have the ability to limit number of mesh hops like in ZD. I occasionally see SmartMesh inappropriately organize into a 2-hop network in this house (basement -> first floor -> second floor), which results in around 30% worse measured throughput compared to the 2nd floor AP directly connecting to the root AP. Fortunately, using the CLI and setting "mesh-uplink-selection dynamic" results in the mesh eventually correctly reorganizing. I've seen the same behavior in ZD, but forcing a hop limit helpful.
• Unleashed seems to have longer boot/reboot times compared to ZD and ZD-managed AP's. The Unleashed Master requires about 2 minutes to reboot before it starts serving clients. The Unleashed meshes take around 1 minute from reboot to serving clients. It seems like both wait something like 30 seconds (cloud provisioning?) in the blinking power light phase, and there's a longer DIR blinking (provisioning) stage compared to standalone. Not a huge deal, but it's something to keep in mind for mesh networks with few roots. I found with ZD, the root AP reboots in less than 30 seconds, so you can often reboot it without clients noticing the network went down. With 2 minute reboots, it is far more noticeable.

Other Observations

• BeamFlex and Ruckus hardware shows an even more dramatic advantage in real world environments. For fun, I had a Netgear R7000 (Wave 1, 3x3:3) consumer AP cross flashed to DD-WRT, which is regarded as one of the best consumer/prosumer performing APs manufactured around the same time as the R500/R600. The R7000 was able to deliver slightly higher speeds compared to Ruckus at 5ft line-of-sight, but through one floor or layer of drywall its performance quickly plummeted. Ruckus was able to deliver 200+mbit observed throughput through 2 walls. With mobile clients, upload performance sucked with the R7000 at any distance, but was stable with Ruckus at similar distances to download. It matches my general observation that Ruckus seems to maintain a huge lead in Rx sensitivity, which is really important with mobile clients, which tend to have low Tx power both to save power and because people tend to hold mobile devices in a manner that covers the antennas.
• Compared to 802.11n, it seems like with 802.11ac, AP placement (especially for meshing) has a much greater relative impact. I assume this is due to the higher MCS rates for 802.11ac requiring extremely good SNR. With the ZF7982's, I was seeing mesh throughput between 75mbit and 100mbit when fine-tuning placement choices. But between a pair of R600's, I was noticing that moving the MAP just a few feet meant the difference between 150mbit and 250mbit, mainly by avoiding a kitchen cabinet of dishes between the MAP and the RAP one floor below. 
• Just sayin', it would be awesome if Ruckus would consider a dual-5GHz or triple-radio AP for intermediate mesh points. Strapping two AP's together to form a Mesh-eMesh pair is both costly and not very visually appealing. I would rather see the ability for an AP offering that can simultaneously serve clients and relay to backhaul without sharing airtime.

Overall, thumbs-up to unleashed! It is a highly attractive offering that should be a no-brainer for SMBs and even appealing to many prosumers / home users. Not sure if Ruckus has been taking note, but between Eero, Luma and others, there's been a lot of expensive consumer AP's hitting the market that primarily promise higher speeds and mesh reliability. IMO, Ruckus still leads the market on both counts.