Contention in the upwards direction.

Does contention happen in the uplink direction as well as the downlink?
Does the one direction being contended imply the other will be too?

--
http://inquisitor.i.am/ | private.php?do=newpm&u= | Ian Stirling.
---------------------------+-------------------------+--------------------------
Money is a powerful aphrodisiac, but flowers work almost as well.
-- Robert A Heinlein.

Ian Stirling wrote:

> Does contention happen in the uplink direction as well as the downlink?


Yes.


> Does the one direction being contended imply the other will be too?
>


Not necessarily, though there is always a minimum amount of error
checking reverse traffic on any link. Therefore this will get delayed if
the reverse channel is very congested.

A lot depned on teh fine tuning of teh TCP/IP stack - larger window
sizes will utilise bandwidth better under congestion if teh channel is
relatively error free, but will degrade rapidly if packet loss happens.

This is basically how traffic 'slows down' rather than 'gets corrupted'
under heavy congestion. Packets are queued in a buffer, and hence
delayed. If a sender has sent a wedge, and got no acknowledgements, it
waits until it gets some. If it never gets any, it waits some more, and
resends the data.

This is why under severe congestion, things can go into meltdown with
everybody resending data.

In addition the routres can, and will, be programmed to queue different
things in different ways, to help shape traffic flow and prevent
catastophic slowdowns. i.e. you might decide that ACK packets should be
given top priority, and delay data packets. 'real time' packets like
voice or streaming video might be given priority as well.

engineers at Cisco spend a lot of time playing with stuff like this and
publishing learned papers. Which BT et al then don't read :-)

A whle suite of options has been put in just to deal with the fact the
Microsoft can't write networking code as well, but that is another story :-)




>

Very true but you forget one important factor.

Look at the Cisco reports and you'll see that the overwhelming majority of
Cisco business is not in the carrier sector, but in the Enterprise area.
Cisco are certainly to be admired but they have largely failed to make any
real inroads into the telco sector.

Good products they have in abundance but even their engineering staff are
very light on carrier grade network experience. BT have far more experience
in the best way to handle hundreds of thousands of users simultaneously.

Despite all Cisco efforts, not a single network relies on their technology -
for two reasons.

Their support is nowhere near good enough for carrier networks and they
expect to do software upgrades every month - just not possible on a
nationwide carrier network.

Secondly, reliability. When the IT gear fails, pick up the phone and call IT
support. Carriers need reliability as good as the telephone and Cisco
generally don't get close. Cisco certainly don't as their software is flaky
at best for the first year - they deploy too early for carrier grade
reliability.

So Cisco produce learned papers, but BT don't read them as the papers aren't
entirely relevant to carrier grade networks.

And yes I know there are Cisco routers at the BT core, however they are
backed up by bigger Foundry Network boxes which ensure availability.


On 11/8/03 12:24 pm, in article (E-Mail Removed), "The Natural
Philosopher" <(E-Mail Removed)> wrote:

> engineers at Cisco spend a lot of time playing with stuff like this and
> publishing learned papers. Which BT et al then don't read :-)

Graham in Melton wrote:

> Very true but you forget one important factor.
>
> Look at the Cisco reports and you'll see that the overwhelming majority of
> Cisco business is not in the carrier sector, but in the Enterprise area.
> Cisco are certainly to be admired but they have largely failed to make any
> real inroads into the telco sector.


Well I am afew years out of date, but for IP routing most IS_P's used
bug cisco 7500's or better. The ones I bu8lt for certainly did. No
reliability problems at all - generally uptimes in teh year plus area,
and then usually only downed to reload configs, not because they faled.


>
> Good products they have in abundance but even their engineering staff are
> very light on carrier grade network experience. BT have far more experience
> in the best way to handle hundreds of thousands of users simultaneously.


Hmm. Voice/SDH yes, IP packets - no. BT have probably the worst ISP
buisiness of all teh UK ISP's. Or did 5 years ago. Only about 3 blikes
understood it.


>
> Despite all Cisco efforts, not a single network relies on their technology -
> for two reasons.
>


Many networks rely on their technology at the IP level. Mine does for
one. :-)


> Their support is nowhere near good enough for carrier networks and they
> expect to do software upgrades every month - just not possible on a
> nationwide carrier network.
>


Extraordinary. I never used to upgrade the cicos at all unless teher was
a spscific problem that needed fioxing,nor did anyone else.


> Secondly, reliability. When the IT gear fails, pick up the phone and call IT
> support. Carriers need reliability as good as the telephone and Cisco
> generally don't get close. Cisco certainly don't as their software is flaky
> at best for the first year - they deploy too early for carrier grade
> reliability.


Cisco software is probably teh best I gahve ever dealt with: new
featires doid often have bugs, but you did';t pick u teh phone - you
used your key into their support center, and chances are it was a known
problem with a known worlaround or patch available. Cisco support from
my perespective was better than anybodies.

Ok the didn't wipe your arse for you - but they showed you where to get
the paper...


>
> So Cisco produce learned papers, but BT don't read them as the papers aren't
> entirely relevant to carrier grade networks.
>
> And yes I know there are Cisco routers at the BT core, however they are
> backed up by bigger Foundry Network boxes which ensure availability.


????

I can't see how you can back up one make of box with another, or if teh
foundries are so effin good, why they aren't used in teh fisrt place.

Anyway, BT is the last ISP on earth I would use as an example of good
prcatice.


>
>
> On 11/8/03 12:24 pm, in article (E-Mail Removed), "The Natural
> Philosopher" <(E-Mail Removed)> wrote:
>
>
>>engineers at Cisco spend a lot of time playing with stuff like this and
>>publishing learned papers. Which BT et al then don't read :-)
>>
>

Not that I'm one to talk, but use a spellchecker !

On 11/8/03 10:30 pm, in article (E-Mail Removed), "The Natural
Philosopher" <(E-Mail Removed)> wrote:

> Graham in Melton wrote:
>
>> Very true but you forget one important factor.
>>
>> Look at the Cisco reports and you'll see that the overwhelming majority of
>> Cisco business is not in the carrier sector, but in the Enterprise area.
>> Cisco are certainly to be admired but they have largely failed to make any
>> real inroads into the telco sector.
>
>
> Well I am afew years out of date, but for IP routing most IS_P's used
> bug cisco 7500's or better. The ones I bu8lt for certainly did. No
> reliability problems at all - generally uptimes in teh year plus area,
> and then usually only downed to reload configs, not because they faled.
>
>
That¹s why I said carriers not ISP's. Uptimes of a year are nowhere near
good enough for single boxes !

>>
>> Good products they have in abundance but even their engineering staff are
>> very light on carrier grade network experience. BT have far more experience
>> in the best way to handle hundreds of thousands of users simultaneously.
>
>
> Hmm. Voice/SDH yes, IP packets - no. BT have probably the worst ISP
> buisiness of all teh UK ISP's. Or did 5 years ago. Only about 3 blikes
> understood it.
>
Same comment - I distinguish between ISP's and carriers. ISP's can only
provide a service if the carriers provide backhaul that basically is 100%
available. I agree BT Openwoe are pretty awful. That¹s not due to lack of
knowledge of IP, but of an inability to run a cost effective operation.
>
>>
>> Despite all Cisco efforts, not a single network relies on their technology -
>> for two reasons.
>>
>
>
> Many networks rely on their technology at the IP level. Mine does for
> one. :-)
>
>
Sorry to harp on, but same point. Carrie scale networks ,no, Small
enterprise scales, yes.
>> Their support is nowhere near good enough for carrier networks and they
>> expect to do software upgrades every month - just not possible on a
>> nationwide carrier network.
>>
>
>
> Extraordinary. I never used to upgrade the cicos at all unless teher was
> a spscific problem that needed fioxing,nor did anyone else.

That¹s because you never bought carrier scale boxes - just normal gear. Have
a look at Marconi BFS Core switches, or the Foundry Networks equivalent -
things that cross connect 480Gb data streams and you'll find Cisco are
nowhere. That¹s a total of 480Gb,up to 1.92 Tb. Cisco can't even get through
the initial testing stage.
>
>> Secondly, reliability. When the IT gear fails, pick up the phone and call IT
>> support. Carriers need reliability as good as the telephone and Cisco
>> generally don't get close. Cisco certainly don't as their software is flaky
>> at best for the first year - they deploy too early for carrier grade
>> reliability.
>
> Cisco software is probably teh best I gahve ever dealt with: new
> featires doid often have bugs, but you did';t pick u teh phone - you
> used your key into their support center, and chances are it was a known
> problem with a known worlaround or patch available. Cisco support from
> my perespective was better than anybodies.
>
Missed the point I think - to build a carrier scale network, you simply
can't have software bugs, and if you do, not very often. The problem with
Cisco is exactly as you say, they know about the bugs and that¹s a symptom
of not writing it properly in the first place. If you've deployed say,
50,000 routers in a carrier network, the last thing you want from Cisco is a
monthly bugfix. It costs too much to deploy and that why no carrier has
deployed their technology on any scale. About 4 years ago, Cisco "announced"
that they would build the next Telia IP network. Telia cancelled the
billion dollar contract as soon as they saw how little Cisco knew about
reliability and availability on a carrier scale network.

> Ok the didn't wipe your arse for you - but they showed you where to get
> the paper...
>
>> So Cisco produce learned papers, but BT don't read them as the papers aren't
>> entirely relevant to carrier grade networks.
>>
>> And yes I know there are Cisco routers at the BT core, however they are
>> backed up by bigger Foundry Network boxes which ensure availability.
>
>
> I can't see how you can back up one make of box with another, or if teh
> foundries are so effin good, why they aren't used in teh fisrt place.

Bit are bit bytes are bytes and the BT IP network is highly redundant. Of
course data streams are backed up - you don't have to use the same supplier
just to stream data ! BT do not rely on a single supplier for anything so
every core IP node is a dual redundant configuration which in some central
nodes actually means two separate suppliers to avoid systemic faults.
>
> Anyway, BT is the last ISP on earth I would use as an example of good
> prcatice.
>
See point about Carrier scale networks and ISP's above

>>
>> On 11/8/03 12:24 pm, in article (E-Mail Removed), "The Natural
>> Philosopher" <(E-Mail Removed)> wrote:
>>
>>
>>> engineers at Cisco spend a lot of time playing with stuff like this and
>>> publishing learned papers. Which BT et al then don't read :-)
>>>
>>
>
>

dave Stanton wrote:

>>I can't see how you can back up one make of box with another, or if teh
>>foundries are so effin good, why they aren't used in teh fisrt place.
>>
>>Anyway, BT is the last ISP on earth I would use as an example of good
>>prcatice.
>>
>>
>
> Is there something wrong with your keyboard ?
>


Loose nut on it somewhere :-)


> Dave
>
>
>
>
>