50M : Think Broadband Ping Monitor Results (POST YOURS)

[Chrysalis]

that user is on a different port anyway I think.

[qasdfdsaq]

Originally Posted by boroboi

Indeed that one has me stumped... could it be that the Router MAC was registered to a different part of the network to the modem MAC? I'm not sure how VM routes it's customers, the channels are the same though.

In router mode, i'd see maybe 18-19ms min. latency. In modem mode it appears to be around 14-15ms from what i can guess looking at the graph, with a marked improvement in jitter, although it's still there.

Your latency already changed before switching to modem mode, and you haven't provided any direct comparison between the two (i.e. before/after/before again). It could have been completely coincidental, something to do with the IP change, or the switch itself.

Looks like your Superhub Normal graph stops around 11:30pm. Your modem mode graph doesn't start until 4pm. We don't even know how many days apart that is, and anything could have happened in the intervening time that's nothing to do with either the Superhub or modem mode.

Comparing two things many hours/days apart really doesn't give us much to go on...

(If you want to compare latency and jitter in the two modes, get a decent ping program, and ping the nearest node you can on the VM network, and do an immediate comparison)

[Sephiroth]

Qasi - are you up on the theory as to why modem mode might have lower latency than router mode? Or, of course, definitely wouldn't have lower latency?

I can see you're saying above that the particular results posted by Boroboi aren't of a standard from which deductions can be drawn.

But, in theory ......

[qasdfdsaq]

It's quite possible that modem mode would have slightly lower latency, but I wouldn't expect that much simply between the two modes. There's a lot of variables though, for example the Superhub's firmware may simply not be programmed to respond to pings at the highest priority, something that is quite common on carrier network equipment. Another example variable would be the SH needing to inspect the IP packet in router mode, determine if it is for itself or a device being NAT'ed behind it and then determine where to pass it to, a process that takes time, and something it doesn't have to do in "modem mode". Also, higher average CPU load in router mode, etc.

In the end I'd expect ~0.5-1ms difference between the two modes but larger differences are probably a result of different routing in other parts of the network.

If we wanted to make a direct comparison between the SH in modem and router mode, best thing would be to ping from something *behind* the SH rather than relying on the SH itself to respond to pings - i.e. use the same endpoints in both situations, and also to try obtain the SH's WAN MAC and clone that if so that the PC/router would get the same IP address from VM's network. That'd rule out any differences in routing based on IP address.

[Sephiroth]

A lot of what you say makes sense. But if you displace the routing functions to an external device, the same factors arise.

I suppose someone could frig around with powerful and less powerful external routers which would confirm or otherwise the postulations you have offered.

[qasdfdsaq]

If we shift the routing to an external device, or remove it entirely (i.e. don't use a router) then we isolate the effect of the Superhub's routing functionality - isn't that what we want?

(I'll also add it's not how much power you have but how you use it, i.e. the quality of the software running on the device is more important. The Superhub runs some funny real-time operating system called eCos. Apart from the name, I know very little about it.)

[boroboi]

Originally Posted by qasdfdsaq

Your latency already changed before switching to modem mode, and you haven't provided any direct comparison between the two (i.e. before/after/before again). It could have been completely coincidental, something to do with the IP change, or the switch itself.

Correct, it did, but even in Modem mode it is a couple of MS faster than (strangely) it was in Router mode if you look real carefully

Looks like your Superhub Normal graph stops around 11:30pm. Your modem mode graph doesn't start until 4pm. We don't even know how many days apart that is, and anything could have happened in the intervening time that's nothing to do with either the Superhub or modem mode.

The two graphs weren't measured at the same time, but were merely for visual purposes to describe what i was seeing. Here are the two time lapsed graphs from when i switched to modem from router... as you can see, the avg latency and jitter itself is noticably lower and/or stable at all times of the day.

There is around a 30 minute difference in the times as it was close to midnight, so i just showed the graph from the next day onwards.

Comparing two things many hours/days apart really doesn't give us much to go on...

As i say, it was just to show a visual representation of what i was describing, nothing to make an in depth analysis over

(If you want to compare latency and jitter in the two modes, get a decent ping program, and ping the nearest node you can on the VM network, and do an immediate comparison)

Indeed that would be the best solution. I'm not really looking to do an in depth analysis, if it looks better on the TBB graph that's good enough for me

[roughbeast]

I'm really quite unhappy about the loss of my mid-evening latency bumps. I've got nothing to moan about.

[img]Download Failed (1)[/img]

Do you think I've had a visit from the latency fairy?

[boroboi]

Originally Posted by roughbeast

I'm really quite unhappy about the loss of my mid-evening latency bumps. I've got nothing to moan about.



Do you think I've had a visit from the latency fairy?

Don't jinx yourself fella

[qasdfdsaq]

Originally Posted by boroboi

Correct, it did, but even in Modem mode it is a couple of MS faster than (strangely) it was in Router mode if you look real carefully

Yes, but because your latency was changing by itself irrespective of router/modem mode, it's hard to concretely attribute any further changes to modem/router mode.

There is around a 30 minute difference in the times as it was close to midnight, so i just showed the graph from the next day onwards.

Given those circumstances, then it's more likely (but not with absolute certainty) that there's only two possible causes - different routing due to different IP, or the modem/router mode switch itself. I suspect a combination of both, as neither alone should really be causing that big a change.

[boroboi]

Originally Posted by qasdfdsaq

Yes, but because your latency was changing by itself irrespective of router/modem mode, it's hard to concretely attribute any further changes to modem/router mode.

The min latency has been up and down twice over the last couple of weeks, but never less than it was in the "dip down" you see in that graph, which is where it has been the most, around 18-19ms, never as low as it is in modem mode currently 14-15ish

[qasdfdsaq]

Originally Posted by Sephiroth

I suppose someone could frig around with powerful and less powerful external routers which would confirm or otherwise the postulations you have offered.

I did a little frigging around.

Superhub router mode, LAN side, firewall off:
RTTs of replies in ms: min/avg/max/dev: 0.322 / 0.363 / 1.091 / 0.037
Superhub router mode, modem side, firewall off:
RTTs of replies in ms: min/avg/max/dev: 0.363 / 0.401 / 1.170 / 0.033

Superhub router mode, LAN side, firewall on:
RTTs of replies in ms: min/avg/max/dev: 0.419 / 0.472 / 1.012 / 0.054
Superhub router mode, modem side, firewall on:
RTTs of replies in ms: min/avg/max/dev: 0.370 / 0.417 / 5.862 / 0.179

Superhub modem mode:
RTTs of replies in ms: min/avg/max/dev: 0.260 / 0.293 / 1.103 / 0.043

VMNG300 modem mode:
RTTs of replies in ms: min/avg/max/dev: 0.338 / 1.321 / 46.820 / 4.460


Not a lot of difference, router mode increases latency by about 0.05ms.

And just a bit of trivia:
A slower router:
RTTs of replies in ms: min/avg/max/dev: 0.298 / 0.328 / 0.765 / 0.021
A faster router:
RTTs of replies in ms: min/avg/max/dev: 0.215 / 0.254 / 0.578 / 0.029
BT FTTC Modem:
RTTs of replies in ms: min/avg/max/dev: 0.387 / 0.467 / 1.279 / 0.053

---------- Post added at 23:17 ---------- Previous post was at 23:06 ----------

Originally Posted by roughbeast

I'm really quite unhappy about the loss of my mid-evening latency bumps. I've got nothing to moan about.

Do you think I've had a visit from the latency fairy?

No, maybe a visit from the network upgrade fairy

[buckleb]

My IP address changed recently, and there is a slight difference.

With 'New' 77.102.x.x IP Address




With 'Old' 94.169.x.x IP Address


Minimum latency has increased, but I don't notice any real difference in real-world use (I don't play ping-dependant games).

110Mb Product

[Jayster]

Can anybody explain what on earth is happening in regards to those peaks and troughs? Those large average and max spikes are downloads while the packet loss seems to be a problem throughout my area as the core and ubr also show it my guess is congestion to tbb somewhere.

[craigj2k12]

its due to the way the VMNG300 buffers packets, basically a lot better than the superhub