Community discussions

MikroTik App
 
User avatar
kiler129
Member
Member
Topic Author
Posts: 352
Joined: Tue Mar 31, 2015 4:32 pm
Location: IL, USA
Contact:

[Discussion] Is MT treating non-PtP wireless seriously?

Sun Sep 26, 2021 7:48 pm

This isn't going to be a rant, I don't want to upset or undermine anyone but simply share my thoughts on the topic and maybe spark a (hopefully constructive) discussion here.

MikroTik hardware and software is amazing. I think most of us really enjoy using it. However, I believe there's an elephant in the room, and I hope it's not just me seeing it. The wireless side of things is just .... bad. Back in the days when I was amazed by the NV2 and how relatively effortlessly links could be brought up. Nobody was thinking to use MT for customers and we were simply terminating it with a RJ45 jack. Times have changed - people expect good wireless for besides PtP links. Unfortunately I believe MT is severely lacking here.


PtMP
I tested a range of devices: some old ones like 2011, some newer ones targeting prosummers like 4011, and things like hAP ac3, wAP ac, hAP ac etc... with a single device and a careful tuning I was cable to get a decent performance. Usually... unless the environment was noisy. Recently I hit an obstacle which I didn't solve - wireless VR requiring a stable high performance link. The budget wasn't an objection: it just needed to work. After swapping multiple devices I gave up and used a non-MT AP - with automatic settings it just worked. Was it more expensive? Yes. Was the client happy? Yes.


CAPsMAN & MPtMP
Ok, the single AP and multiple clients scenario usually works ok but as soon as we need to add multiple APs things get.... ugly. CAPsMAN is a great piece of software but it's still not a central management solution as many things have to be done manually on every AP. Even a simple thing like updating software: I can update ROS on a group of APs but... not the firmware. But this isn't the big problem. The lack of intelligence is.
CAPsMAN - not undermining its efforts - is just a configuration synchronization and a tunnel. All radios seem to be living in a vacuum without awareness of their surroundings. This is bad, I will even call it horrible for moving clients performance. Sure, we use reject lists to force reconnection between APs, we follow the purists idea of clients picking the best AP. However, this just doesn't work, clients are dumb. What broke the camel's back for me was a customer with a large house (a typical American home, just large but with paper-thin walls). Six APs were working wonders as long you were connected to the correct AP. However, clients notoriously were sticking with the AP or even connecting to a lower-signal AP causing 2Mb/s throughput. Reject access list based on signal wasn't a solution - obviously it helps but doesn't solve the problem.


Hardware limitations?
I wonder if this is an issue with hardware or the software lacking here? ROS is amazing for routing but feels dated in the wireless space. I'm not talking about 802.11ax, that is a separate problem. I'm talking about the "secret sauce" which seems to be missing here. I really, really, REALLY don't want to mention players like ruckus here, but the truth is I can toss 1-2 of their APs and they will just work communicating magically between each other giving clients great performance. Are they 3-4x more expensive? Sure. But how much is my time worth attempting to fine-tune all settings and positions... which will be thrown out of the window when the client moves a metal cabinet with documents, and how much I will be even able to do?


Is it just me?
I wonder is my growing frustration with wireless on MT is me getting older and more grumpy, or things really need a solid shake-up?
 
gotsprings
Forum Guru
Forum Guru
Posts: 2087
Joined: Mon May 14, 2012 9:30 pm

Re: [Discussion] Is MT treating non-PtP wireless seriously?

Sun Sep 26, 2021 10:58 pm

You typed all that, like you didn't know that Mikrotik wireless is ACv1.
 
User avatar
kiler129
Member
Member
Topic Author
Posts: 352
Joined: Tue Mar 31, 2015 4:32 pm
Location: IL, USA
Contact:

Re: [Discussion] Is MT treating non-PtP wireless seriously?

Mon Sep 27, 2021 1:20 am

You typed all that, like you didn't know that Mikrotik wireless is ACv1.
Wave2 or not.... even without 802.11k/v/r (which isn't the solution by itself) the regular 802.11ac is inferior.
 
User avatar
anav
Forum Guru
Forum Guru
Posts: 18958
Joined: Sun Feb 18, 2018 11:28 pm
Location: Nova Scotia, Canada
Contact:

Re: [Discussion] Is MT treating non-PtP wireless seriously?

Mon Sep 27, 2021 2:09 am

Rip Van Kiler ??
Sorry bud your as late to the discussion as MT is to providing a fully capable wifi 5 device.
I have been using the tplink eap245 instead of my capacs for some time now.
 
mducharme
Trainer
Trainer
Posts: 1777
Joined: Tue Jul 19, 2016 6:45 pm
Location: Vancouver, BC, Canada

Re: [Discussion] Is MT treating non-PtP wireless seriously?

Mon Sep 27, 2021 3:45 am

I'm hoping that we might see these types of features once wifiwave2 is up and running. I agree they are necessary. I've been running into similar problems with roaming, and letting the device decide is not good enough.

MikroTik's decision to make their own wireless drivers instead of using the manufacturer's drivers puts them at a disadvantage in today's market. I'm sure there were good reasons for doing this initially, probably related to being able to fit drivers for many chips on devices with a small flash. Other manufacturers didn't have to be concerned with this, because the OS was only for that device and only needed to include the driver for that one wireless chip and not 10-20 different wireless chips. Because RouterOS is somewhat monolithic, and it is a single RouterOS npk for all devices of a particular architecture, the package has to include drivers for all of the wireless chips, not just the wireless chip in that model. As a result you have "wasted space" in the npk for drivers that are not even needed for the particular model that you have.
 
User avatar
kiler129
Member
Member
Topic Author
Posts: 352
Joined: Tue Mar 31, 2015 4:32 pm
Location: IL, USA
Contact:

Re: [Discussion] Is MT treating non-PtP wireless seriously?

Mon Sep 27, 2021 11:24 pm

Rip Van Kiler ??
Sorry bud your as late to the discussion as MT is to providing a fully capable wifi 5 device.
I'm struggling... really. I'm trying to keep the stack simple and central as the fewer device types to support (especially running different stacks) the easier it is to manager. The COVID situation cranked the problem to 11. Before that it was manageable and I was simply complaining here and there nudging people around. Now in the other hand I just feel like I had enough. This is just my 5 cents. We know that MT reads forum - they don't always respond but they see it.

Today I had another "wifi is working poorly" type of a call. Of course another site with multiple APs. I think iOS 15 tweaked things again with how it roams between APs.
I have been using the tplink eap245 instead of my capacs for some time now.
I think I have 2-3 of them. They're... ok. Certainly better than what MT offers but nothing to write home about.

I'm hoping that we might see these types of features once wifiwave2 is up and running. I agree they are necessary. I've been running into similar problems with roaming, and letting the device decide is not good enough.
Wifiwave2 changes nothing.... sorry. It's still a dumb driver living in the vacuum of its own ignorance. Adding insult to injury wifiwave2 doesn't work on RB4011... or I should say it does work for 5Ghz only removing the ability to use 2.4Ghz all-together. This is seriously not acceptable. If I deployed an RB4011 to a power user I cannot give them a choice "do you want better WiFi for your laptop without your IoT or maybe do you want your IoT to work?". 2.4Ghz, as bad as it is with interference, is still widely used. Sure, MT will probably release RB5009 with dualband supporting wifiwave2 but still today RB4011 is the flagship with WiFi for endpoint deployments where powerful CPU and multiple ports are desired.

MikroTik's decision to make their own wireless drivers instead of using the manufacturer's drivers puts them at a disadvantage in today's market. I'm sure there were good reasons for doing this initially, probably related to being able to fit drivers for many chips on devices with a small flash. Other manufacturers didn't have to be concerned with this, because the OS was only for that device and only needed to include the driver for that one wireless chip and not 10-20 different wireless chips. Because RouterOS is somewhat monolithic, and it is a single RouterOS npk for all devices of a particular architecture, the package has to include drivers for all of the wireless chips, not just the wireless chip in that model. As a result you have "wasted space" in the npk for drivers that are not even needed for the particular model that you have.
I think making their own driver was an amazing choice... in pre-ac times. They really had an edge over other solutions. NV2 was great but I think they aren't able to keep up (or don't have people?) with the vastly more complex ac featureset. I doubt CISCO, Arguba, Ruckus or other big players who do WiFi REALLY well use just vanilla drivers from the SoC vendors. We also have to remember that times were different - now the "radio" is no longer a radio.

I think there are two major problems:
  • throughput & link stability => this is probably solvable with wifiwave2
  • interdevice communication
The second one is big. It's not [just] about a single pane of glass management - this is less important and can be solved with automation tools. The issue is that APs cannot live in their own vacuum and rely on manual tuning. It simply doesn't scale when an average joe in the office nowadays carries their phone, smartwatch, and a tablet while actively transiting data walking around on their facetime audio call.
 
mducharme
Trainer
Trainer
Posts: 1777
Joined: Tue Jul 19, 2016 6:45 pm
Location: Vancouver, BC, Canada

Re: [Discussion] Is MT treating non-PtP wireless seriously?

Mon Sep 27, 2021 11:53 pm

Wifiwave2 changes nothing.... sorry. It's still a dumb driver living in the vacuum of its own ignorance.
It does. I think you misunderstood me. I know that by itself, it doesn't handle anything regarding roaming and the 802.11 r/k/v standards. But with the current drivers, MikroTik is having to reinvent the wheel, and is spending so much time doing that that they would never have the time or manpower to build an 802.11 r/k/v solution. With wifiwave2, at least the driver is there, already made by the manufacturer, which then MikroTik shouldn't have to do much with. This frees up developers time to build a new capsman version for wave2 with 802.11r/k/v support.
 
User avatar
kiler129
Member
Member
Topic Author
Posts: 352
Joined: Tue Mar 31, 2015 4:32 pm
Location: IL, USA
Contact:

Re: [Discussion] Is MT treating non-PtP wireless seriously?

Tue Sep 28, 2021 2:43 am

Wifiwave2 changes nothing.... sorry. It's still a dumb driver living in the vacuum of its own ignorance.
It does. I think you misunderstood me. I know that by itself, it doesn't handle anything regarding roaming and the 802.11 r/k/v standards. But with the current drivers, MikroTik is having to reinvent the wheel, and is spending so much time doing that that they would never have the time or manpower to build an 802.11 r/k/v solution. With wifiwave2, at least the driver is there, already made by the manufacturer, which then MikroTik shouldn't have to do much with. This frees up developers time to build a new capsman version for wave2 with 802.11r/k/v support.

Ok, true that - I give you this. However, I wonder why they had the manpower to build solutions like NV2 or now they do rather complex things like L3 offloading on so many platforms... it's like they had good radio engineers and they don't anymore. While they take to blob from the manufacturer they still need to integrate a lot. I reallly hope they will be able to bring more of the smartiness to the market. It's not about CAPsMAN v2 - it needs a lot more than just k/v/r.
What I'm worried about is that they're now dependent on the manufacturer: they use a blob and a config file (which we all saw in another thread). They can't (correct me if I'm wrong) modify the driver source as they don't have it. It's a similar problem as Raspberry Pi had - they had to go own silicone route as broadcomm refused to give them the sources for the GPU blob.
 
mducharme
Trainer
Trainer
Posts: 1777
Joined: Tue Jul 19, 2016 6:45 pm
Location: Vancouver, BC, Canada

Re: [Discussion] Is MT treating non-PtP wireless seriously?

Tue Sep 28, 2021 4:26 am

They can't (correct me if I'm wrong) modify the driver source as they don't have it. It's a similar problem as Raspberry Pi had - they had to go own silicone route as broadcomm refused to give them the sources for the GPU blob.
I read online that Qualcomm does provide vendors like MikroTik with the driver source code for the wireless drivers, and that the source code itself is GPLv2 licensed. When I have asked MikroTik about missing features in wifiwave2 like four address mode support, they responded "not available yet", rather than something that is not possible and will never be available with wifiwave2. Both imply that MikroTik has the ability to make their own changes to the Qualcomm drivers to facilitate the use of certain features. I can't see them adding something like NV2 on wifiwave2 as then they would probably have to open source their NV2 code, but for everything else, it may very well be possible.

Who is online

Users browsing this forum: Amazon [Bot], grusu and 34 guests