[Gllug] SwitchedOnUK

[James Roberts]

Simon Wilcox wrote:
> On Thu, 16 Nov 2006, James Roberts wrote:
> 
>> Jason Clifford wrote:
>>> In practice IP telehony is very reliable these days. You'd only get
>>> latency issues if you have a poor IP connection. BT have plenty of
>>> bandwidth and a good network to push this over.
>> Hah! You *have* to be kidding! :)
> 
> I doubt he is, Jason knows quite a bit about these things (clue - check
> his sig :)

So do I :)

>> Given a 2Mbps SDSL 10:1 inter-office line, and a good backbone provider,
>> we cannot reliably provide 30 users with connectivity interoffice
>> (average use 10 lines at a time). This is *supposed* to support 50 users
>> at the two office with 2Mbps out at teach end.
> 
> What else goes over that connection ?

Nothing. Nada. Null. Rien. :) Data uses a separate line (also crap).

> At 10:1 contention that means you may at times only get 204kbit/s. I'd
> class that into the "poor" category.

Thank you for the reminder :) It's easy to forget the basics :)

> You need a 1:1 contention if you're running lots of concurrent calls.

Never would have believed it! :D

> Are you running any QoS on the link, esp if you;re running anything else
> over it.
> 
>> We have told out client to go back to pots. Alternative: leased line for
>> VoIP. Economies? Hah!!
> 
> Compared to a leased line for inter-pbx private traffic then yes as it can
> be shared with other data traffic or at least a backup as part of your
> business continuuity provision.

With the greatest respect: financial calcs (guesses?) here roughly 
equivalent to talking out of poopah, sorry.

Cost pcm of old POTS ISDN system (with 100% uptime)

	say £1450

Cost current system

	say £1100
	(but unreliable - client has lost significant multiple K work)

Cost with leased line

	say £1900

Show me the saving if it is moved to leased line :)

We have 15 client workstations running remote desktops served by Citrix 
on the other line BTW.

In terms of fallback connectivity: no cable available at either end, no 
other supplier. Have set up 3G/GPRS/UMTS fallback but that's only 370K up.

>> This system is supplied by a major VoIP supplier, who say 'none of our
>> other clients suffer from this'. But I have talked with some of the
>> 'other clients' and they are *not* in accord with this view.
> 
> Name and shame then. Who is the backbone provider and who the major VoIP
> supplier. If they're rubbish then we should be told !

Backbone: Mistral

VoIP - can't. Commercial confidence is commercial confidence. However, 
you can guess around things that are devilish with numbers after them.

The problem is nothing to do with the backbone provider BTW. Our NOC
  is showing good figures to the routers at each end of the link.

The problem is with BT. The exchanges at each end are crap. The one in N 
London is un-staffed: they have to parachute engineers in from 
Hertfordshire and half of them don't seem to know what they are doing. 
They go to fix one line and un-patch the other - more than once now. 
When it *is* patched its performance varies at random. Getting 
performance reports from BT is not easy. The SLA for SDSL that is on 
offer gives a 72 hr window before they have to escalate priority on repairs.

The inner London exchange is bad as well; though does have staff 
engineers on site.

Now we have not seen performance as unreliable as this elsewhere: we are 
a reseller for Mistral and have plenty happy clients.

SLAs can be upgraded: Leased Lines can be installed. But the LOGIC of 
VoIP is to save money. For this client it has not done so, it has cost 
them money and clients (try 3 days of no phone service and 3 months of 
unreliable service), and making it as reliable as the old pots will cost 
more per month than putting the pots back. See the problem?

Incidentally, it was *not* our idea for them to go VoIP :)

> Do think they've tested it

Doubt it :(

They don't care.

Look: current technology is designed to guarantee the delivery of a 
packet in a timeframe. IP is not designed to do this. QOS on IP does not 
replace the guaranteed packet delivery. VoIP on IP v6 with inbuilt QOS 
does not guarantee (AIUI) delivery in the same timeframe as current 
technology does.

VoIP works adequately because most calls are not using it, they're on 
pots. If/when the whole global voice backbone moves to IP then I see a 
few issues ahead. Obviously doing so has tremendous cost benefits for 
the main connectivity providers... who are losing money to the parasitic 
companies running VoIP over the spare bandwidth over the backbone. So 
why maintain spare bandwidth on the backbone?

Well, that's my view...

With respect :)

Please correct my errors :D

JR
-- 
Gllug mailing list  -  Gllug at gllug.org.uk
http://lists.gllug.org.uk/mailman/listinfo/gllug