MU-MIMO

Moo-MIMO

I readily admit that I could be all wet on this blog, so I have posed the title as a request rather than an assertion. I am going to state my current opinions on various aspects of 802.11ac Wave-2 technology, as it relates to enterprise customers, in the hopes that if I’m off-base, someone smarter than me will help me understand the error(s) of my thinking. If receive no arguments, or instruction, regarding the information listed below, then I will make the (hopefully correct) assumption that my current opinions are correct.  I’m going to go through a variety of Wave-2 related topics and list my thoughts – in no specific order.

Transmit Beamforming (TxBF)

* No WiFi Alliance certification and no certified equipment

* High channel overhead due to channel matrix feedback for every participating client.

* Not needed at short range (because RSSI is already plenty high) unless you’re trying to do MU-MIMO (which has even more overhead).

* Not good at long range because the channel feedback matrix is too complicated.

* Can’t use with Spatial Multiplexing

MU-MIMO

* Requires TxBF support, which isn’t available.

* Downlink technology only, so even the theory isn’t that great.

* No WiFi Alliance certification and no certified equipment

* No client support (either for MU-MIMO or TxBF)

* Downlink traffic for orthogonally-positioned devices is unlikely to simultaneously fill AP queues unless every device in an area is streaming unicast simultaneously.  This would only happen in niche use cases.

160MHz Channels

* 160MHz channels. Never, ever….EVER…use 160MHz channels in the enterprise.

* 80MHz channels. No. Don’t. Not even “dynamic” 80MHz channels, because they cause a CCI mess in enterprise environments. Been there, seen that. It’s horrible.

* 40MHz channels. No. Stop it. Unless… they are only used in very specific areas for very specific (and strongly justifiable) reasons. Use of ubiquitous 40MHz channels detracts greatly from system-wide capacity and greatly increases the chance of CCI with nearby systems. With small networks, where there aren’t many (or any) neighbors, you might be OK using 40MHz channels, but otherwise, NO.

* Enterprises use 20MHz to maximize system-wide capacity and to minimize CCI. Do this.

* Hint: enable 40MHz Intolerant so that your neighbors can’t use wide channels either. 

4 Spatial Streams (4SS)

* No clients available with 4SS capability.

* Eventual 4SS clients will only be desktops and large laptops, which are in the extreme minority due to today’s highly mobile society.

256QAM

* Great when you are within arm’s reach of the AP and can easily cable at 1Gbps full duplex, but even the best APs in the market, when used with any average client device, won’t be able to sustain 256QAM at anything over 50 feet (~16 meters). You can often expect client/AP connections to downshift modulation from 256QAM to 64QAM even at <30 feet.

* Requires very high SNR, which depends on a low noise floor.  Noise floors are unpredictable and rarely changeable.

* The high RSSI needed can easily cause CCI in the network design, so its not a good idea to design for 256QAM.

Receive Sensitivity

* Wave-2 chipsets may have better Rx sensitivity than Wave-1 chipsets (or they may not), but that doesn’t help anything. In fact, it can be a hindrance given that most networks are poorly designed and have high CCI. The Rx Sensitivity difference between Wave-1 and Wave-2 chipsets would be miniscule anyway, so this is a “who cares” issue.

Better MRC

* With 4 receivers, an AP can hear better, so perhaps there will be less uplink retries, but most clients use too much power already (because they aren’t controlled by the infrastructure and don’t comply with 802.11h TPC), and a 4th receiver won’t add enough to matter in most real-world use cases.

PoE

* It’s already been proven that 4×4:4 can fit into 802.3af (standard PoE) but most high-end Wave-2 APs will operate on 802.3at (PoE+). It’s very unlikely that any Wave-2 APs will exceed PoE+, so PoE itself, in regard to Wave-2, is a non-issue.

Backhaul

* Under no enterprise circumstances, will you need more than 1Gbps for Wave-2 APs. #ThatIsAll

Client Types

* Most clients are either 1×1 or 2×2 11n or 11n/ac, and the bulk of the rest are legacy 11a/b/g crap. In what real-world use case (scenario) will Wave-2 help?

Throughput

* Throughput for a given BSS depends on the number of clients, type of clients (e.g. 1×1, 2×2), applications and QoS ACs in use by each client, simultaneous transmissions, instantaneous data rates of each client, interference (WiFi and non-WiFi), and so much more. An AP’s capabilities are in the minority of affecting parameters. Wave-2 capabilities would have little to no appreciable benefit in any reasonable use case.

Design

* Most WiFi networks are poorly designed, configured, and rarely ever tested. Wave-2 APs have no chance of adding value with a poor design, and due to the reasons above, add minimal or no value even with a good network design.

Data

* There is no vendor-specific or vendor neutral data showing real-world throughout benefits of MU-MIMO.

DFS

* If Wave-2 APs don’t yet support DFS channels due to not yet having FCC (or other regulatory authority) certification, they actually have LESS capacity than older APs due to less available channels. You especially need to remember this in very high density (VHD) use cases.

Future

* Don’t expect to see any 6×6 or 8×8 systems delivering client access, as 4×4 has already hit a ceiling of diminishing returns. The pricing of 4×4 systems are already at a shocking premium, but adding more powerful DSPs and double the number of radio chains is unreasonable for the foreseeable future.

Pricing

* The only real value that I see Wave-2 is that with the introduction of Wave-2 APs, there is a corresponding reduction in the price of Wave-1 APs. I now strongly suggest to my customers that they consider opting for mid-range 3×3:3 Wave-1 APs (very cost effective, while still having significant processing resources), and if they are itching to spend extra money, they should buy additional Wave-1 APs that will be deployed as dedicated performance/security sensors.

What To Do

* If you want a much faster network, don’t buy what you think are faster APs (when they aren’t), but instead, do the following:

1) Move to 5GHz: lock, stock, and barrel. Don’t use 2.4GHz at all.  2.4GHz is dead.

2) Get rid of old client devices, now. They are a major performance drag.

3) Increase your minimum basic rate to 24Mbps. Disable all rates below 24Mbps.

• Yes, I know that that would prevent you from using 802.11b – see item #2 above.

4) Use 20MHz channels.

5) Design your network for an RSSI of around -64 to -67dBm with minimal CCI.

6) Get rid of as many RF interference sources as possible.

7) Use only APs that support DFS channels, and replace client devices that don’t support DFS.

8) Minimize your SSID count. 4 is the max. 3 is better.

9) Use dedicated sensors rather than background scanning, and constantly monitor your network for performance-impacting issues.

10) Other stuff:
http://divdyn.com/top-ten-tune-tips

An excellent Q&A on 802.11ac Wave-2 featuring Matthew Gast is 802.11ac Q&A. I agree with Matthew’s points, across the board.  He did a good job clarifying several salient points in this podcast.

That’s it. Please help me understand where my current thoughts are off-base.

------------

http://divdyn.com/moo-mimo/