20k PPS * 353Byte average iMix packets = 56,480,000 bits. That jives.
On 10/23/2020 9:32 AM, Matt Mangriotis via AF wrote:
The original 450 AP and SM were roughly the same PPS, and had similar
limits… the AP having more FPGA gates, it has some additional
resources to handle more things. In practice, in a typical network
with mixed packet sizes, this ends up around 55-60 Mbps in a 20 MHz
channel (PPS did improve over time going from 13k when they were first
released, to ~20k these days… but we reached the limit of what those
chips could do).
The good news is that 450i, 450b and the new MicroPoP radios are all
built using a next generation FPGA (SoC), which has embedded ARM
processors, and with Release 20, we’ve unlocked some serious gains in
PPS on these bad boys… effectively increasing PPS by 250% or more
(from ~40k to >100k). If it’s a budget-constrained site, I might
suggest a MicroPoP as an upgrade, depending on how many SMs you need
to serve (and how far away they are).
Standards-based chipsets have a distinct advantage in this area
because the MAC/PHY is baked into the ASIC chip and does what it does
very well… but limits the flexibility on what can be done. With our
current approach, nearly anything is possible given time and
resources, because everything down to Layer 1 is coded into that FPGA…
but it adds a bit of cost.
Matt
*From:* AF <af-boun...@af.afmug.com> *On Behalf Of * Ken Hohhof
*Sent:* Friday, October 23, 2020 8:21 AM
*To:* 'AnimalFarm Microwave Users Group' <af@af.afmug.com>
*Subject:* [ External ] Re: [AFMUG] 450 and 450i AP CPU limit on
throughput?
Not talking about sustained speed keys (finally gone in the 450b) but
the point at which you haven’t maxed out the RF capability but the CPU
horsepower becomes the limiting factor and you can’t get any more pps
through it. I think the SMs also have this issue, but less likely to
be a problem unless using them in PTP mode or having a small number of
very high bandwidth customers.
Or maybe you are saying the APs never had the CPU limitation like the
SMs, but I’m pretty sure they did. Each generation 450, 450i, 450m
having a more powerful CPU. Although not as powerful as we might
think from the price, I guess maybe the result of using a processor
core in an FPGA, it seems like lowly WiFi chips have more CPU power. I
think most of us were surprised to find the limitation could be CPU
not RF.
*From:* AF <af-boun...@af.afmug.com <mailto:af-boun...@af.afmug.com>>
*On Behalf Of *Adam Moffett
*Sent:* Friday, October 23, 2020 6:42 AM
*To:* af@af.afmug.com <mailto:af@af.afmug.com>
*Subject:* Re: [AFMUG] 450 and 450i AP CPU limit on throughput?
I thought the limitation was on what a single SM did.
.....maybe mistaken.
On 10/23/2020 12:14 AM, Ken Hohhof wrote:
Does anyone know off the top of your head what the current CPU
limited max throughput is for 450 vs 450i APs? Not based on RF
characteristics but packet processing in the CPU.
I keep thinking at one time 450 APs were only capable of maybe 20
Mbps but that can’t be right because I have some doing over 40. I
think Cambium said that firmware tweaking kept raising that number.
I’m asking because I have one lone site with a 450 and an omni,
and maybe a dozen subs scattered through the entire 360 degrees.
So while I am going to need more throughput, it just doesn’t
justify 4 sectors, and the cables are in conduit and I think we
only ran 4 cables and we have 2 backhauls. And it occurs to me
what I need isn’t sectors, it’s to increase the channel width to
30 MHz (450) or 40 MHz (450i). But will a 30 MHz channel really
help if the 450 AP is pps limited by the CPU? I’m OK with
replacing the 450 with a 450i if necessary. Most sites we have at
least 4 sectors, so mostly 20 MHz channels. But an omni with a 40
MHz channel would use the same amount of spectrum as 4 sectors and
20 MHz channels.
--
AF mailing list
AF@af.afmug.com
http://af.afmug.com/mailman/listinfo/af_af.afmug.com
