[Article] - Is MU-MIMO Ready For Prime Time?

[sfx2000]

On the main site

http://www.smallnetbuilder.com/wireless/wireless-features/32898-is-mu-mimo-ready-for-prime-time

I'm not surprised at the data/results - did the modeling some time back, and with current chipsets, 3 MU clients is likely the maximum, and there's other issues that still need some work.

Tim's testing with the new test bench confirms my suspicions...

MU has it's place, but there's a lot of heightened expectations that might be dashed...

 

[RogerSC]

Wanted to thank Tim for doing these tests. Good to see what's going on with this technology that's been out there in theoretical form for quite a while. Real interesting.

Thanks!

 

[RMerlin]

Also, one thing that just came to mind here: mobile devices seem to be going for dual streams. My Nexus 5X and Nexus 9 are both 866 Mbps devices. Same with most 802.11ac-enabled laptops so far. Makes me wonder even more how useful MU-MIMO will actually be when the vast majority of clients will be dual, not single stream.

 

[iwod]

1. You need at least two MU-MIMO devices to get any benefit from MU-MIMO

Why? What happen if I have a One MU-MIMO devices and 1/2 Non MU-MIMO devices? Would the router just turns lowest common and uses SU-MIMO for all devices?

2. MU-MIMO works only for downlink data (moving from router to device). It provides no benefit for uplink

That is for 802.11ax. Which is stupid, we are having to incrementally buy upgrade sort of like every year. I also wonder do upload stops the downloads in 802.11ax ( i.e Half Duplex? )

I kind of expect if there are lots of client, MU-MIMO will improve throughput, turns out we dont get anything. But do latency improves?

 

[htismaqe]

Also, one thing that just came to mind here: mobile devices seem to be going for dual streams. My Nexus 5X and Nexus 9 are both 866 Mbps devices. Same with most 802.11ac-enabled laptops so far. Makes me wonder even more how useful MU-MIMO will actually be when the vast majority of clients will be dual, not single stream.

My iPhone 6 is the same way. It always grabs two streams if it can.

 

[red_pope]

I read the following reviews to get a better understanding of Mu-mimo technology. My conclusion is my personal opinion.

Most current readings:
http://www.smallnetbuilder.com/wire...mu-mimo-ready-for-prime-time?showall=&start=1

Most oldest readings -

http://www.cisco.com/c/en/us/soluti...rks/802-11ac-solution/q-and-a-c67-734152.html
http://www.smallnetbuilder.com/wire...s-xstream-the-coming-battle-for-wi-fi-airtime
http://www.webopedia.com/TERM/M/mu-mimo.html
http://www.windowsnetworking.com/ar...less-networking/mu-mimo-vs-su-mimo-wi-fi.html
http://chimera.labs.oreilly.com/books/1234000001739/ch04.html#beamforming_basics
http://www.intechopen.com/books/recent-trends-in-multi-user-mimo-communications
http://www.radio-electronics.com/in...input-multiple-output-technology-tutorial.php
https://en.wikipedia.org/wiki/Multi-user_MIMO

Bottom line in my opinion, the technology is there.
But lacks abundance and availability of hardware.
Lacks Better Firmware, research and development from the manufactures.

Yes, my eyes are hurting from reading and my brain cells are bursting. I need my brandy with honey! then my dog will be rename " el mimo musu", ill bet you he will urinate my door steps when he finds out.

Good Job everyone

 

[thiggins]

1. You need at least two MU-MIMO devices to get any benefit from MU-MIMO

Why? What happen if I have a One MU-MIMO devices and 1/2 Non MU-MIMO devices? Would the router just turns lowest common and uses SU-MIMO for all devices?

MU-MIMO increases total throughput by enabling up to three devices to share the same airtime. Only MU devices can do this. So one MU and one SU device will require two transmit airtime periods. No throughput gain.

2. MU-MIMO works only for downlink data (moving from router to device). It provides no benefit for uplink

That is for 802.11ax. Which is stupid, we are having to incrementally buy upgrade sort of like every year. I also wonder do upload stops the downloads in 802.11ax ( i.e Half Duplex? )

MU-MIMO works using a special form of transmit beamforming. So it works downlink only.

I kind of expect if there are lots of client, MU-MIMO will improve throughput, turns out we dont get anything. But do latency improves?

Not true. The tests show there is a gain in total throughput for the Linksys and NETGEAR products out to 10 clients or so. The drop below SU throughput for 4 devices is I hope, an anomaly that will be fixed.
I did not look at latency, so don't know.

 

[RadarJammer]

MU-MIMO increases total throughput by enabling up to three devices to share the same airtime. Only MU devices can do this. So one MU and one SU device will require two transmit airtime periods. No throughput gain.....

Here, those routers with two 5 GHz radios would make really sense, if I get it right: one used exclusively for MU clients, one for SU ones.

 

[sfx2000]

Here's part of the challenge for MU-MIMO, cribbed from my notes a while back...

 

[sfx2000]

Here, those routers with two 5 GHz radios would make really sense, if I get it right: one used exclusively for MU clients, one for SU ones.

No, not really... the SU clients are going to get fair airtime in any event - it's really the MU clients that will have some challenges... and many implementations might say "what the heck" and drive the packets SU and still leverage TxBF in any event...

Depends on the traffic...

 

[avtella]

I know Quantenna is sampling an 8x8 ac chipset but after hearing about all the issues with their previous implementaion, I hope Qualcomm or Broadcomm also come out with their own variants quickly. As Merlin said earlier with newer mobile devices coming out with dual WiFi antennas it would be better to have MU-MIMO with more than 4 antennas.

 

[sfx2000]

MU-MIMO increases total throughput by enabling up to three devices to share the same airtime. Only MU devices can do this. So one MU and one SU device will require two transmit airtime periods. No throughput gain.

Stumbled across this a while back that gives a great illustration of what MU can do for overall channel efficiency - now note that individual clients aren't going to be much faster (actually not faster at all), but certain scenarios can make the link available more often for SU clients..

 

[thiggins]

sfx: Could you please cite the source for the diagrams?

I don't think it was ever claimed MU devices would have higher PHY rates. The claimed advantage has always been higher total throughput.

 

[sfx2000]

Cisco...

http://www.cisco.com/c/en/us/soluti...rks/802-11ac-solution/q-and-a-c67-734152.html

 

[avtella]

I believe Qualcomm mentioned an overall 200+ Mbps hit with MU-MIMO in throughput due to overhead and "training packets" and that it could even be worse depending on the situation but still overall beneficial.

This seems like a pretty good article:
http://semiaccurate.com/2015/04/27/qualcomm-talks-802-11ac-mu-mimo-details/


It seems from the Qualcomm roadmap the IPQ 8065 does not suffer from the n-1 MU-MIMO streams problem and I believe Broadcom also mentioned about working on solving that problem in another article which I can't seem to find. Anyone have thoughts on that?

 

[paraplu]

Currently using a MUMIMO Samsung S7 as client to my sitecom Greyhound router (same Qualcomm hardware as tplink ac2600).

Question: is it normal that max linkspeed is 433 in MUMIMO mode, compared to 866 in SUMIMO?

 

[thiggins]

Currently using a MUMIMO Samsung S7 as client to my sitecom Greyhound router (same Qualcomm hardware as tplink ac2600).

Question: is it normal that max linkspeed is 433 in MUMIMO mode, compared to 866 in SUMIMO?

2x2 MU-MIMO STA is a valid configuration. But can't say whether it is normal. Could be Samsung or chip vendor steps down to 1x1 for MU mode.

 

[sfx2000]

Currently using a MUMIMO Samsung S7 as client to my sitecom Greyhound router (same Qualcomm hardware as tplink ac2600).

Question: is it normal that max linkspeed is 433 in MUMIMO mode, compared to 866 in SUMIMO?

The AP makes the decision to either do SU or MU, and it determines the link rate/stream config based on feedback from the MU and SU clients, and current traffic loads...

Totally reasonable that it might consider a MU single stream frame over a SU two stream frame based on overall feedback from all clients...

(MU depends on feedback not only from MU clients, but all clients within the BSS that speak 802.11ac, not just MU, but also SU)

 

[sfx2000]

I don't think it was ever claimed MU devices would have higher PHY rates. The claimed advantage has always been higher total throughput

I agree - MU has always been about capacity and link utilization and overall throughput, even though any specific STA is still limited to what it can do, and in many cases, it might be lower to a specific STA, but generally... it depends on the entire BSS - how many clients, how many can do MU, what the traffic load is across the clients, and any upstream (Router/AP) prioritization might come into play...

Scheduling is hard - whether or not to do MU, does the beam forming matrix support that instance when it might work, and how to recover when it doesn't.

Qualcomm (outside of Atheros, and combining Atheros works as well) and Broadcom are probably at the forefront for this in 802.11 at the moment - it's deep and serious work, and scheduling algorithms are very proprietary - and one can appreciate the challenge that Quantenna is facing, as they do not seem to have a client chipset in wide distribution to develop/test against...

Interoperability is going to be the toughest nut to crack - consider Intel, who has a wide customer client base, but has to test against QCA/Broadcom, and at some point Quantenna and Marvell (and sometime in the future, MediaTek/Ralink and Realtek if they decide to do the MU thing...)

MU is a very big challenge from both a AP and Client perspective - I wouldn't hold one's breath for a rapid deployment of this feature in 2016...

 

[thiggins]

The AP makes the decision to either do SU or MU, and it determines the link rate/stream config based on feedback from the MU and SU clients, and current traffic loads...

True. Could it also be possible that the STA's firmware makes a decision to do MU in 1x1 only also?