Greetings:
I understand this likely depends on the capabilities of the hardware at both ends, so let's say we are talking about the latest supported Ubiquity Rocket and Mikrotik BaseBox models on a sector antenna.
With 10Mhz channels, how many >15dB connections can one of these nodes support before the node starts to fall over? Said another way, how many RF connections can a node reasonably support before the mesh starts to suffer?
How does your answer depend on traffic? For example, a bunch of nodes that can see the central node but are just doing AREDN station-keeping traffic, vs some number of nodes taking to different set of nodes through the one central node? Do things start to degrade as soon as one node starts talking to another node, or can we have N connections in-flight before the central node starts to max out?
Thanks,
Charles
I understand this likely depends on the capabilities of the hardware at both ends, so let's say we are talking about the latest supported Ubiquity Rocket and Mikrotik BaseBox models on a sector antenna.
With 10Mhz channels, how many >15dB connections can one of these nodes support before the node starts to fall over? Said another way, how many RF connections can a node reasonably support before the mesh starts to suffer?
How does your answer depend on traffic? For example, a bunch of nodes that can see the central node but are just doing AREDN station-keeping traffic, vs some number of nodes taking to different set of nodes through the one central node? Do things start to degrade as soon as one node starts talking to another node, or can we have N connections in-flight before the central node starts to max out?
Thanks,
Charles
Reference: How many RF connections is too many?
Hi, Charles:
Short answer: 999. ;-)
Medium answer:
A. Chances are very good that it will take several years for a new network build to get close to 'too may RF connections'.
B. Chances are very good that builders of a new network will add poor network designs that will inhibit its performance
which will render non-relevant a concern about 'too many RF connections'.
Long answer: All of the reference material I have read is concerned with maximum throughout,
whereas your concern is maximum number of RF connections.
Reduced throughput:
A 10 MHz channel at 15 dB SNR may have 1/40 the throughput of a 20 MHz channel at 26 dB SNR.
Reduced channel availability:
A 10 MHz channel may take up to twice as long to transmit the same data as a 20 MHz channel.
A 15 dB SNR channel may take up to twice as long to transmit the same data as a 18 dB SNR channel.
A 15 dB SNR channel may take up to 20X as long to transmit the same data as a 26 dB SNR channel.
Strive to achieve no hidden nodes. https://en.wikipedia.org/wiki/Hidden_node_problem
Strive to achieve no exposed nodes. https://en.wikipedia.org/wiki/Exposed_node_problem
Strive to have fewer stations sharing the same band-space. i.e. Avoid 'mesh' topology.
'node starts to fall over' 'mesh starts to suffer'
Short answer: When the 3rd node is added to a channel.
Long answer:
'mesh' is a topology, not a unit of measure of network efficiency.
Your frequent use of 'connections' reminds me of packet radio (AX.25) connections/connected-state.
AREDN is using Wi-Fi 'ad-hoc' mode and its radio circuits might be better regarded as connection-less. ?
I defer referencing 'current' devices, but when I joined our local AREDN network 5+ years ago,
there were 6 high-profile nodes and 26 lesser profile nodes all on the same channel and the network looked
good on a http://localnode.local.mesh/cgi-bin/mesh page.
However, the throughput was so miserable, you could not view any nodes' status page except your own!.
Today we 60+ nodes supporting 50+ services and use multiple channels on 5 GHz to link.
We still have active 2.4 MHz high-profile node switches, but no users remain active on that band.
I hope this helps, Chuck
If there are two stations on channel, each gets half the bandwidth. If there are 10 stations on channel, each gets 1/10th the bandwidth.
One way to optimize your throughput is to adopt the cellular method: A moderately high access point node services an area on channel A. Another moderately high access point node services a different area on channel B. They're linked together by another pair of nodes on channel C (which doesn't necessarily have to run AREDN, since it's not intended for user access. (This is basically what Chuck said
Obviously there's more equipment needed to do this, but in exchange you'll see greatly improved throughputs.
Orv W6BI
Thanks, that helps a lot! For some reason I didn't think that a 3dB signal difference could cut transmission time in half. And the cellular method definitely is making much more sense (other than the extra expense/maintenance).
Best,
Charles W1ZPB