Superhub : Can Someone Explain WHY the SuperHub is so...bad?

[Kymmy]

Closest you'd get is pinging the UBR but even then it would be dependent on how utilized the unit was

[Skie]

Which is why you do a large number of real-world tests of both devices. With a big enough dataset you can take into account the odd variances and come up with a decent estimation.

Nobody has done this though :p

[Kymmy]

"real world tests" will not tell you the latency.. it'll just tell you the state of the route especially as on long haul routes the latency might be less than 1% of the overall figure

With extremely short routes the latency might be as much as 50%+ so you can discern if latency is having a detrimental effect

[Peter_]

Originally Posted by Skie

Which is why you do a large number of real-world tests of both devices. With a big enough dataset you can take into account the odd variances and come up with a decent estimation.

Nobody has done this though :p

So every single speedtest to Wellington on either the VMNG300 or the Superhub follows the exact same routing over 12,000 plus miles each and every single time.

If you believe that for one second then I have this bridge for sale on a one day special offer for a great price to you only.

[zekeisaszekedoes]

Originally Posted by Stephen

They don't need to be technically adept, as most of VMs customers probably aren't. The technically minded are more than likely a minority in this case.

Not all of them do, granted, but it stands to reason that the techie types are going to test the CPE in ways regular customers might not, plus they can articulate their experiences more accurately. If the idea is to get things fixed as quickly as possible then getting the more skilled people involved seems like the most logical choice.

IMO what you've just said is exactly the flawed VM thinking that I've referred to in the past, i.e. not testing properly enough and wondering why people are getting ticked off with unreliable kit. Makes no sense at all... completely illogical, as Spock once said.

Originally Posted by KenK

He has already - #22.

I missed that one, but usually in threads like this he says it once or twice anyway.

Originally Posted by Kymmy

What I'm suggesting is that unless you can ping the exact same site from the exact same location using the two different set-ups (superhub and non-superhub) then the result will be skewed by the differing states of the existing traffic/route for the difference in either time or location.

So you're recommending we have 2 CPEs on our account to test from exact same location. Um... you do know that's against the VM terms of service, right?

[Kymmy]

Originally Posted by zekeisaszekedoes

So you're recommending we have 2 CPEs on our account to test from exact same location. Um... you do know that's against the VM terms of service, right?

I never said such a thing.. I just pointed out that it would be the only way of getting a plausible result

Anyway at one point I had both a 2050 and a 255 running on the same cable even though it was only for a few hours when I switched from VM to VMB

[Chrysalis]

Originally Posted by Kymmy

Only if you can do the pings both at exactly the same time is the result relevant The internet traffic along said route can change from second to second changing your results..

ok I see the position is to be obstructive then thanks for making it clear.

I know all about changing conditions and such but internet routes if not congested do not change from second to second by more than a few ms.

I also always do repeat tests so if I was to post 2 traceroutes you can be sure it is something I have done at least 3 or 4 times and the post would be representative of the average at that moment.

By saying I or others cannot post traceroutes or ping times showing the affect the superhub has on the internet I feel CF has lost it and is no longer a neutral forum.

Saying things like traceroutes cannot be trusted because of changing conditions etc. is the sort of thing an isp would say to defend a poor network.

Even VM's own forums dont apply this much censorship to the superhub. Banning users, and forbiding traceroutes been posted.

Of course VM's network which is congested does have changing conditions which is the exact reason I will do multiple tests, do you really think that the fact me and others after using both devices for months and my graphs showing obvious increases of latency is talking rubbish and the network just happens to change its behaviour every single time I change the device, maybe you suggesting there is some guy flicking a switch everytime chrysalis puts in the superhub?

For your information I have used traceroute's alongside other tools to diagnose network problems for many years, this is the first time someone has tried to discredit in such a fashion of 15 years of using the internet.

---------- Post added at 14:44 ---------- Previous post was at 14:28 ----------

here is some data, not a traceroute so not gone against a mod.



Did VM flick a switch or did the network suddenly change behaviour when I swapped the modem

[Skie]

Originally Posted by Masque

So every single speedtest to Wellington on either the VMNG300 or the Superhub follows the exact same routing over 12,000 plus miles each and every single time <snip immaturity>

Which is why you do a large number of tests. The more results you have the more realistic your data set is and the better the accuracy you will have when you come to baseline the data.

You can never get a perfect test method, but there are ways to mitigate these sort of things. Or did you lot skip statistics lessons in school?

But like I said, nothing anyone has done or is likely to do will be thorough enough to get a definitive answer as to how much of a ping increase the Superhub is adding to a connection over the VMNG300, if any. We just have to go on the limited results here.

[Peter_]

Originally Posted by Skie

Which is why you do a large number of tests. The more results you have the more realistic your data set is and the better the accuracy you will have when you come to baseline the data.

Once the routing goes outside of the Virginmedia network the is no way to ensure that every traceroute or speedtest follows the exact same path as it will always choose the shortest and fastest route so if the is heavy traffic elsewhere it will automatically reroute, so as you cannot control the exact route it one takes how can you be sure of any accuracy due to these inconsistencies.

[Skie]

Originally Posted by Masque

Once the routing goes outside of the Virginmedia network the is no way to ensure that every traceroute or speedtest follows the exact same path as it will always choose the shortest and fastest route so if the is heavy traffic elsewhere it will automatically reroute, so as you cannot control the exact route it one takes how can you be sure of any accuracy due to these inconsistencies.

You dont need it to be consistent. If you do 10k pings over a week then you have a good baseline. You mitigate the routing inconsistencies by having such a huge sample size that it is irrelevant as to where the packets are going via and the only really consistent thing is the startpoint (which we are interested in) and destination. Do another 10k over a week with another modem and then you have your data.

[Peter_]

Originally Posted by Skie

You dont need it to be consistent. If you do 10k pings over a week then you have a good baseline. You mitigate the routing inconsistencies by having such a huge sample size that it is irrelevant as to where the packets are going via and the only really consistent thing is the startpoint (which we are interested in) and destination. Do another 10k over a week with another modem and then you have your data.

Pinging a website over 12,000 miles away is never going to give you consistent results.

[ethan103]

Originally Posted by Masque

Pinging a website over 12,000 miles away is never going to give you consistent results.

I always get the same ping?

[markie1966]

Originally Posted by Masque

Pinging a website over 12,000 miles away is never going to give you consistent results.

well ive just let winmtr run for an hour or so to a game server that i play on...surely thats consistent enough for u?? have a look at the packet loss

its bad enough that im losing 3 or 4 % locally....but the next 4 nodes are all on the vm network...ranging from 42% packet loss up to 92% packet loss

is it any wonder that some people are complaining?

i ask you once again...will this new revision or update for the superhub resolve my issues with the vm network as regards gaming...ie no packet loss for smooth gaming??? and if you say it will...can i have it in writing please??

[markie1966]

sorry....heres the winmtr image

[qasdfdsaq]

Originally Posted by Chrysalis

why would it only affect 1 hop? there is nothing special about the gateway.

Because the Superhub acts as a network bridge between two segments of network. One is your LAN, the other is the DOCSIS link between your CM and the gateway. Anything beyond that is dealt with by VM's core network, which in theory would act in exactly the same way regardless of whether the data came from a VMNG300, Superhub, or an Ambit modem.

Beyond the single (logical) hop DOCSIS link your data is bridge back onto a (hopefully) all IP core network again and the Superhub loses any influence it may have as it no longer directly signals anything on this section of the network.

---------- Post added at 20:56 ---------- Previous post was at 20:55 ----------

Originally Posted by Kymmy

Only if you can do the pings both at exactly the same time is the result relevant The internet traffic along said route can change from second to second changing your results..

Ping -t.

Averages.

Balance of probabilities.

It's not hard.

---------- Post added at 21:01 ---------- Previous post was at 20:56 ----------

Originally Posted by Kymmy

Which is the whole point... You can't measure latency accurately over an ever changing medium...

Yes you can, as long as your definition of "accurate" is not 100.0%. There are acceptable levels of accuracy and standard deviations. Granted, you can't ever be 100% accurate, then again you can't 100% accurately state you're alive or your modem is connected. The best scientific knowledge we have is not 100% accurate.

95% or 99% accuracy is the baseline standard for pretty much all modern sciences, and is more than acceptable for everyone else. Stop being pedantic.