Netonix Forums

I see detection of outage as being the true delay in all this.

What would be great is if we could get some radio support for something like BFD. The radio, or rather radio's OS will have detailed statistics on how the link is performing where all the link agnostic solutions are inferring the link conditions. We do need both, but having information from the radio would improve detection times.

You could identify a link's declining performance by knowing how many frames were re-transmitted or how often error correction is being used and if that is changing over time.

If the routing protocol would begin transmitting packets tagged as latency priority via a backup route as soon as the primary link started to show signs of trouble (pre-failure) then convergence times could be typically be eliminated except when a link hard-fails.

This all comes down to a clearer view of the physical layer though.

This is what the Performant gear did.... I don't know why they didn't succeed =\

mhoppes wrote:This is what the Performant gear did.... I don't know why they didn't succeed =\

Did they fail or get bought out?

also, people are very unforgiving for proprietary stuff. Having a big network tied to a single vendor that isn't omnipresent is risky. build an RFLO network and then they go out of business and you have to replace that gear for instance....

if an RFLO like protocol were open and available and modular so that various radios could have a shim daemon to feed the brain information, I think it would be adopted by wISPs readily. assuming you could run it on various hardware, say ERLites.... or something affordable.

Ummmm... it is a standard.
http://en.wikipedia.org/wiki/Ethernet_R ... _Switching

mhoppes wrote:Ummmm... it is a standard.
http://en.wikipedia.org/wiki/Ethernet_R ... _Switching

well, RFLO is/was based on that with proprietary adds and inoperability with any other vendor.. so based on standards or not it was effectively propprietary yes?

It was yes, but the standard exists if you want to run it... I just wish UBNT (OR CHRIS!!!!) would build it into their switches.
*cough* Chris? Please please please?

mhoppes wrote:I just wish UBNT (OR CHRIS!!!!) would build it into their switches.

Well... when someone suggested it on the UBNT forum, they were called out as a lone wolf.
http://community.ubnt.com/t5/EdgeSwitch ... 98895#M666

That's because someone *cough*:) Said you can't run more than about 200-300 devices on a layer2 network. Which is completely untrue. That's only if you do it wrong.

There are many major carriers running thousands of devices across a layer2 network. It's all in how you chop up the transport pipe. Yes, you don't let 5,000 devices all ARP to each other. That doesn't mean the transport layer can't be flat back to the router.

At any time between local subs, roaming subs, and broadband we probably have between 5 and 10,000 end user devices running across our flat network. It runs fine all day long.

mhoppes wrote:That's because someone *cough*:) Said you can't run more than about 200-300 devices on a layer2 network. Which is completely untrue. That's only if you do it wrong.

There are many major carriers running thousands of devices across a layer2 network. It's all in how you chop up the transport pipe. Yes, you don't let 5,000 devices all ARP to each other. That doesn't mean the transport layer can't be flat back to the router.

At any time between local subs, roaming subs, and broadband we probably have between 5 and 10,000 end user devices running across our flat network. It runs fine all day long.

haha, thanks for partially quoting my and skipping the relevant part, you know where I said to cache or proxy arp :p

and my point was valid, the demand appears to be very low for this feature. maybe that's because none of the proponents are outspoken.

mhoppes wrote:That's because someone *cough*:) Said you can't run more than about 200-300 devices on a layer2 network. Which is completely untrue...

LOL You don't know how much I had to bite my tongue and sit on my hands to not challenge that statement. I've only been doing networking for about 25 years, so what do I know?