RPi4B does asymmetrical Gigabit with SQM for me. 1000 MBit down but my upstream is capped around 10% of that.
Since x86 was a no go, I'd assume RPi is too...
1 Like
You are right.
Some significant CPU like an ARM SBC or x86 is really the only good way to handle gigabit + sqm.
Archer C7 is a terrible choice and will top out around 200Mbps just routing no SQM if I remember correctly.
Perhaps just buying a slower tier service, maybe 100Mbps or 200Mbps, will be more cost effective and pleasant. A bloated gigabit connection is no fun, and yes it can happen.
1 Like
I think it can go up to 1gbit, w/o SQM.
I use a C7, and no...it can't go up to 1 Gbps...with or without SQM.
1 Like
From:
Don't confuse bandwidth with throughput...totally different things.
And...as mentioned on the wiki page -
Enabling wireless, running additional applications, or even imperfect ISP lines can significantly limit “real world” performance to well below these values.
What is the different? I can't see any meaningful one. Anyway do you have any other suggestion?
You're kidding, right?
I have some resources you might check out -
@frollic gave you the info for 500 Gbps and up connections.
A rewording of this might be "don't confuse theoretical speeds with actual speeds". For example, the C7 has gigabit ports, but realistically it doesn't route anything close to a gigabit per second. I'm not sure what it'll do with software flow offloading, but without it, 200Mbps of throughput is about the right approximate range.
Really if you're getting a gigabit right now, x86 is likely your best bet. RPi4 availability has dried up quite a bit and/or become very expensive.
1 Like
Theoretical vs real-world speeds apply to both bandwidth, and throughput.
Bandwidth and throughput are two different things.
Unclear. Bandwith is a term used in computing/networking quite differently from other fields, and alas in computing it does not have a strict definition and can mean anything from maximum capacity or gross rate to achieved/consumed throughput or goodput. IMHO it is a term that needs a qualification what is actually meant.
3 Likes
Bandwidth is a size or rating or something you design for or aim to achieve.
Throughout is something you measure.
Really if you're getting a gigabit right now, x86 is likely your best bet. RPi4 availability has dried up quite a bit and/or become very expensive.
That's pretty much my conclusion.
My wish is for Gigabit routing with SQM as well. With a a side order of the lowest possible power consumption - so the obvious choice is a Raspberry Pi 4 with two Gigabit Ethernet interfaces: one for upstream, one for downstream, which you plug into a cheap Gigabit switch (preferably VLAN capable) and add as many access points downstream of the switch as you need, placed in optimal locations (which probably won't be where the router is, or the Internet connection to the upstream ISP).
The current problem is that most Gigabit-routing-with-SQM-capable x86 platforms are either expensive or power-hungry or both; and Raspberry Pi 4s are difficult and/or expensive to get hold of, especially in non-mainstream places.
The OPNsense DEC740 looks lovely - 15 watts and no fan, but at 700 Euro is out of my current price range (Note 1).
The DFRobot Raspberry Pi Compute Module 4 solution also looks good, but runs into the problem of lack of availability of the RPi compute module.
Note 1: Electricity prices can make some seemingly expensive equipment relatively cheap. At 10 Eurocents per kilowatt hour, 10 Watts power draw continuously works out to 8.75 Euro per year. Scale up or down according to local average electricity prices and equipment power draw. If you intend to operate your device for several years, it can be worthwhile to pay more for a more power-efficient one.
Bandwith is a term used in computing/networking quite differently from other fields,... IMHO it is a term that needs a qualification what is actually meant.
I agree completely. For old timers, it is much like the definition of baud-rate, which was subject to confusing misinterpretation.
Goodput is what people usually want to know, but can be impossible to calculate or measure accurately. There is a whole industry built on measuring application performance. The IEEE/IEC/OSI probably publish a standard traffic mix somewhere for performance testing of network queueing systems. It is not simple.
Not a bad definition, but alas not the only one out there.
However goodput still tends to be simpler to measure than say the exact gross limit of a link; after all, goodput is the throughput achieved at the "layer" of interest and if that layer is selected carefully it is easy to measure. E.g. IPv4/6-TCP goodput is not that hard to measure (which is why almost all speedtests measure exactly that), IP goodput is already considerably trickier to measure.
There is also the option of using a raspberry pi400 with an USB3 ethernet dongle... I bought one partly for that purpose, but it turned out it makes a sufficiently usable light load desktop machine that I never got around testing its suitability as router...