[TelstraClear] Cable vlan segmentation / ARP reduction

Forums › One New Zealand (including Vodafone, WxC, Farmside) › [TelstraClear] Cable vlan segmentation / ARP reduction

1 | 2 | 3 | 4 | 5 | 6 | 7

33 posts

Geek

#150332 22-Jul-2008 15:22

Quidam:

Anyway, thanks for your reply, and I'll hazzard a direct question.

If am pinging my default gateway, what sort of numbers would you call "good" or "normal"?

typically, I get below 20ms and when the system seems to be running optimally, I'm getting close to 10ms (and sometimes under)

There are times when the pings go all over the place, (eg spikes into the many hundreds;thousands or even a complete timeout)

What would you associate that with? ie. is this a manifestation of ARP flooding or something else (eg congestion)

But I'll understand if you feel you can't be specific or even answer the questions.

It can vary depending on the OS you have (not windows if you're seeing 20ms to the gateway :) ) anything over ~60ms and there's probably "a network event" occuring that we're usually pretty quick to find and fix.

michaeln

238 posts

Master Geek

Trusted

#150340 22-Jul-2008 15:37

doppleganger: I think alot of the confusion about the whole contention ratios comes from Glassmans comments about the TCL node size in Wellington and that there's one cmts per node... Which would be an issue if TCL had 100% penetration... Just because a node has 2000 homes in it (and I'm neither confirming nor denying that, call it a number for a theoretical discussion :P ) it doesn't mean all of those houses have cable.

To put it plainly: the number of households in a 'node' isn't at all the same thing as the number of modems on a CMTS.

There might be many 'nodes' assigned to one CMTS, there might be many CMTS assigned to one node.

On a piece of coax in an HFC (hybrid fibre coax) cable network, modern industry standard practice would be around 500 households passed (HHP), This comes from the electrical characteristics of the coax---the signal loss, number of amps required, target signal to noise ratio etc.. At the point where fibre-optic turns into coax, a standard design is to feed up to four coax legs off one fibre. On that basis, 2000 HHP per 'node' would be a normal standard cable build, but that's nothing to do with the number of cable modems per CMTS.

Note also households passed. The electrical design allows for 100% connection, but in practice there will be far fewer than that. Again, this doesn't tell you anything whatsoever about the number of modems per CMTS.

Quidam

291 posts

Ultimate Geek

#150367 22-Jul-2008 16:12

doppleganger: It can vary depending on the OS you have (not windows if you're seeing 20ms to the gateway :) ) anything over ~60ms and there's probably "a network event" occuring that we're usually pretty quick to find and fix.

Thanks.

In fact I am running Windows: Vista Ultimate 64 bit, on fairly hi spec hardware -Core 2 @ 3ghz+ (with gigabit lan).

And I'm not kidding. When things are running well, I'm get continuous pings close to 10ms and sometimes even lower (I've seen the occassional 8-9ms).

In part that's what made this so frustrating. I went from an awesome connection where I was able to host full rooms (better than all my Aussie friends) to being constantly dropped from every game I joined.

But as you have implied, things are definately looking up :-)

update: Just ran the ping test on my partners low-spec laptop running XP home:

Here's what I got:

Reply from 121.73.26.1: bytes=32 time=10ms TTL=127