AiMesh and daisy chain

[someuser08]

I've been running 2 ax92 in AiMesh for while now connected via 5-2 AX backhaul between 3 floors (nodes at ground and 2nd floor). This gave me maximum 270mbps connectivity (to the internet) from top floor with a signal strength of approximately -80 and usage of 160Mhz band. I think its actually not too bad speed considering the signal drop (I have foil based insulation in the floors, so that's why it drops a lot), but since my internet is now capable of 500mbps I thought I would get a third node to stick in between.

The results are rather disappointing. I daisy chained them floor by floor and even though the signals are excellent now (-59 and -50) I only get 360mbps from top floor. I'm testing using only wired connections and from the first floor I now get full 500mbps, which means the first hop is working perfectly. The signal for the second hop is even better that the first, but am I expecting too much from a daisy chain at these speeds? Considering 360mpbs is maximum I get when routers are in the middle of the room unobstructed and if I hide them out of sight it goes down to 310-320 which is really not worth it for the third router.

P.S. There are also seem bugs in the latest Asus firmware - if a middle node of a daisy chain powered off, it won't came back in the middle until you pin uplink for the third node back to it. And optimization of the network does nothing to create a daisy chain even though the signals between the nodes are very different.

 

[L&LD]

Can you make the main router be in the middle of the two nodes? If not, a third unit is detrimental in your environment. (As I would expect from your description).

 

[someuser08]

No, if I could I would not need a third node at all - single node is enough to cover 2-3 floors wirelessly and I just need maximum wired connection to the second floor where my office is.

I would not say the set up is detrimental completely, but considering the cost of a node and not much improvement in speeds its kind of pointless.

Is there anything out there (by Asus or others) that will give me much faster speeds (>40% more) than ax92 via dedicated backhaul?

 

[L&LD]

Not anything I know when setup, effectively, in Repeater mode.

The cost of the third (and second) router may be better spent running a few cables to where the main router really needs to be?

 

[thiggins]

@someuser08 You've discovered the dirty little secret of mesh systems. You can have great signal level, but lousy throughput.

Even with a dedicated channel for backhaul, that additional node causes a retransmit of the packet, which reduces bandwidth. This article, especially the diagrams in the "Closing Thoughts" section.

Router Or Wi-Fi System: How To Choose - SmallNetBuilder

Trying to decide between a router and a Wi-Fi System? This experiment should help you choose wisely.

www.smallnetbuilder.com

 

[someuser08]

Its obvious that in order to receive and transmit the data at the same time at the same speed you have to have double the bandwidth. For the middle node that would mean 500+500=1000mbps which should be still doable as I saw the actual connection was 1700+mpbs to each side. My experience that actual speed is about 60% of wireless connectivity, so that should still allow to achieve 500on the second node unless there are some other bottlenecks in place (CPU, IO or something like that on the router)...

 

[someuser08]

Not anything I know when setup, effectively, in Repeater mode.

Yes, but I meant really just two node set up with better signal than ax92. For example would two ax11000 do a better jobs (not that I'm going to try)?

 

[L&LD]

The signal isn't the issue.

When the daisy-chained routers are in (effectively) repeater mode, it doesn't matter how strong the signal is to the end client devices.

Did you check out the excellent funnel analogy that @thiggins offered above?

I can't explain it better than that.

 

[someuser08]

May be I'm being think here... I do not see how the article is explaining my particular problem. I do not expect final speed to be more than at any point downstream. Say, we have 3 nodes connected like: node 1 -> node 2 -> node 3.

Speed measured on node 1 (via LAN port) - 500Mbps
Speed measured on node 2 (via LAN port) - 500Mbps, this means funnel is big enough at this stage to achieve full speed. Connectivity stats of the backhaul -59dbm
Speed measured on node 3 (via LAN port) - 360Mbps. Connectivity stats of the backhaul -50dbm, better than previous hop, which means if we was to start with just node two (without it having to receive data via hop 1) it would have been 500mbps speed there as well.

The only difference between Node 1 and Node 2 is the fact that Node 2 has to retransmit signal at high speed, so it seems like hardware/software problem of the router itself. Unless this is expected due to wifi specs and there a formula that allows you to calculate what is possible to retransmit via AX protocol...

I will probably do an experiment at some point by testing nodes in close proximity (within 1m of each other, so funnels analogy will be irrelevant) and I have a suspicion that I will not get a maximum on the third node even then...

 

[L&LD]

The main router, with a dedicated radio for wireless AiMesh backhaul on both the main router and an AiMesh node, can give high performance for the network. Because the 2x, 5GHz radios, can operate concurrently (one for backhaul and one for client devices, wired or wireless, on the node). This is what gives full speed. Signal strength differences of a few dBm are irrelevant because it is in the required 'good enough' range the devices are designed to operate within.

When a second AiMesh node is added, and that node connects to the first AiMesh node (and not the main router), the system effectively behaves like it's in Repeater mode. This means now the first hop from the main router to the first node must wait for the hop from the second node to the first node before it can transmit again. This is causing the performance hit.

Therefore I suggest you consider moving the router to the middle of the three levels and (possibly) getting rid of both AiMesh nodes and getting a better overall network experience too. Use the least amount of Wifi required, and no more, for the best networking experience. A single router that has the coverage required is preferred over multiple units. Unless you can use wired backhaul and place the routers at opposite corners of the home (see the links below).

Your proposed test isn't worth doing. 1m is too close. Routers are designed to be used at greater than 3m distances.

If you want to test what I'm suggesting, set up your test like the diagram below.


AiMesh Node----------------------5m----------------------Main router-----------------------------5m---------------------------AiMesh node.


The idea is that the distance between the nodes is much further than the distance to the main router (for each node).



Repeater mode = wireless AiMesh

AiMesh Ideal Placement


And to update the 'visuals' in the link above to your situation.


AiMesh node-----------------------------------------------------------------------------------
--------------------------------------------------------------------------------------------------
--------------------------------------------------------------------------------------------------
---------------------------------------AiMesh main router-------------------------------------
--------------------------------------------------------------------------------------------------
--------------------------------------------------------------------------------------------------
--------------------------------------------------------------------------------------------------
-----------------------------------------------------------------------------------AiMesh node

 

[someuser08]

Ok, are you saying that irrespectively of the connection speed between the nodes the physical latency of retransmission (at any speed) is causing the performance hit and the ceiling of the bandwidth will be mostly independent of wireless speed connectivity? I.e. whether nodes connect on 500, 1000 or 5000mbps the actual speed will be similar? In this scenario indeed the hardware and placement of nodes would not matter...

Funny about 3m distance - I recon currently my nodes are about 2-2.5 meters apart as they are on top of each other on different floors. But because of floor insulation I get zero signal on 5-1 band two floors apart. Only 5-2 ax and 2.4 are coming through.

I should clarify - i'm not trying to build a good WiFi network. I'm trying to build a good LAN network over WiFi. I have very extensive LAN networks on the ground and second floor (and nothing on the first where only bedrooms are) that needs good connectivity. I can't wire them internally as I'd be going through child's bedroom where I can't afford any mess. And I can't wire externally as the wall is a pitched roof...

The idea to have router on the first floor would have been ok for internet access but not for connecting two LANs...

 

[altacus]

I know I'm late to the party, but have you tried, instead of connecting the ethernet to the 2nd floor node, on the 2nd floor, connect the ethernet to a managed switch? This should allow you to connect the 2nd and 3rd floor nodes both to the switch which is connected directly to the router.

 

[L&LD]

Welcome to the forums @altacus.

Yes, you're late. Very late.