Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[kitz]

Kitz thanks for remind me how bad BT's routing is from the east midlands.

A lot of it will depend how far you are from the RAS and core-nodes. 

As you can see from the trace I posted above, it took me 8ms to traverse 55 miles to Manchester, yet once I was on the core it only took 5 seconds to get from Manchester to Ealing, London which is circa 200 miles...  so you can see how much better the core is than the backhauls.

21CN has more core node access points than the old 20CN network, so it depends just how far east you are, but I thought the midlands was one of the areas were access was supposedly improved as new PoPs were introduced :/

[NewtronStar]

Keep thumping the ground the worm will surface.

[boost]

BE1... is it worth posting your full stats? For thread history, if nothing else, perhaps?


As an aside, I've seen 9ms on ADSL2+ PTM/fastpath which I thought was only possible on FTTC prior!

Capping tones to negate INP sounds like a brilliant idea, too, fairplay! A 5ms ping would be epic! Who cares if usable bandwidth is cut in half

[NewtronStar]

Who cares if usable bandwidth is cut in half

The Kitz forum may be over run with posts with "I think my line is capped" 

[kitz]

Ive seen a few 7ms on the old 20CN, but they were all from the London area.

[Bald_Eagle1]

Unfortunately my connection resynced a couple of days ago & I'm now interleaved again with lower DS sync speed:-

http://forum.kitz.co.uk/index.php?topic=13934.msg262948#msg262948

[Chrysalis]

Kitz thanks for remind me how bad BT's routing is from the east midlands.

A lot of it will depend how far you are from the RAS and core-nodes. 

As you can see from the trace I posted above, it took me 8ms to traverse 55 miles to Manchester, yet once I was on the core it only took 5 seconds to get from Manchester to Ealing, London which is circa 200 miles...  so you can see how much better the core is than the backhauls.

21CN has more core node access points than the old 20CN network, so it depends just how far east you are, but I thought the midlands was one of the areas were access was supposedly improved as new PoPs were introduced :/

There is one new POP which I am not been routed through.

I think the southern east midlands is pretty bad for BTw routing.

BT need some POP's between Leicester and London.

[boost]

No worries.

It sounds as if some of us are on 'joking fastpath' anyway. I wonder if it might be possible to use the two latency paths l keep hearing about?

I reckon I could get by on 256Kb/s of pure fastpath goodness. DLM could interleave/RS/INP/FEC the arse out of the remaining 39Mb~.

I wonder how I'd get my gaming traffic to use this slice of fastpath? Win 7 can set DSCP on a per application basis.
I wonder if any of this is doable. There must be a few low end tones that could be considered globally fastpathable? Every profile could be a gamer-compatible profile!

[Chrysalis]

Yeah, actual speed is dependent on both RTT and capacity of the line.  A higher sync speed doesnt mean will get higher speeds from any specific location.

e.g. if my line synced at say 50mbit on a interleaved profile, I would downgrade to 40/10 to get extra snrm and get the line back to fast path, to me 40mbit fast path beats 50mbit interleaved.

[boost]

I came across a graph t'other day that showed the effect of INP on bandwidth and latency. The added latency is annoying to me personally but if the graphs were to be believed, it has a HUGE effect on actual throughput.

Struggling to see the point of INP permanently applied to a line for a noise burst that happens once or twice a day for 500 micro seconds.

I did see some other stuff about 'coding gain' (not related to INP but to RS or interleaving IIRC) but that went over my head tbh. My takeaway from it was that there can be some additional throughput in some scenarios due to better bitloading (or should that be bitwangling? ) but it was a bit much for me to understand so that might well be wrong!

[Chrysalis]

In reality is probably no point but the issue I guess from openreach side is what if the user is watching netflix at that moment and that short error burst is enough to disrupt it?  It may reduce support callouts by a few % and for their eyes may justify it.

I can understand it if the line is unstable for large chunks of the day and its turned on right away when needed, but I think its silly if a line has a short burst of noise, then interleaving is turned on 12 hours later for no reason.