In reply to:

---

but obviously not everyone will remain logged on at the same time.

---

Even with routers?. I never switch mine off

I'm speaking hypothetically James, but all I'm trying to establish is, if a managed network becomes heavily congested is it possible that some traffic is dropped before being prioritized?.

I think we need to define the funnel first. If the funnel is the core network routers then the funnel opening is effectively all the transit and peering links as well as the cross site links for access to the portals, etc.

The amount of traffic these links can handle is several times the amount that central pipes can handle. For example, there's a 10Gbps link to LoNAP but there's ~4Gbps of central capacity.

So the chance of getting to the point where the funnel is full and it drops traffic before identifying it is very very low simply because of how much extra traffic that actually needs to be.

You then have two funnels within that funnel I guess, the first is the Ellacoya which applies the rate limits and tags the traffic, the rate limits will reduce the amount of traffic and the priorities and then used in the second funnel, the ERXs to allow the higher priority traffic through above the lower and all the rest.

Now don't get me started on the BT network and whether or not that looks like a sieve.

Thanks James and Dave for your time on this. It's definately more technical than I can take in
Basically PN guarantee fast HTTP and gaming by slowing down all non time sensitive traffic. Non managed networks let all protocols compete with each other.

Is it possible to determine what the lag could be between the two on a good day?

I recall a discussion, a couple of years ago, about disconnecting idle connections (after a set time) because there was a finite number of connections that a central could support.

I'm guessing to make it a fair comparison you would have to have the same conditions/loads on identical networks, one with management, one without to see any really differences

Yep.

As per my earlier post, this is 32,000 connections per central.

I'm assuming that "automatic disconnects" was just one of the ideas being tossed around as it's never happened to me and wouldn't have proved very popular, I suspect.

[Edit - Dave answered you, but here's what I wrote before he did]

We need one of Dave's analogies really, but as he's busy I think he'd give you something like this:

Imagine a 50 lane motorway (our connection to the Internet), changing to a normal 3 lane motorway (BTs ADSL Infrastructure), then widening up to a 6 lane motorway (Our customers networks at home).

Because the 3 lane motorway is the bottleneck, there isn't room for all the traffic that wants to get through. Instead of allowing a free for all though through the bottleneck, we have introduced a new by-law along the lines of 'any traffic on which is urgent can travel in a prioritised lane'. Slower traffic is assigned other lanes and we can vary the width of the lanes and choose how much traffic is allowed down each lane at any time to ensure there is always enough room for the priority traffic.

We decide what traffic is allowed to travel in the fast lane before it gets to the motorway and mark it by painting it with a coloured blob. Traffic with the right coloured blob is directed to the appropriate lane on the motorway and there are queues on the entrance to some lanes (but once it's travelling it all goes at the speed limit). Fortunately, IP is better behaved than humans or else this analogy wouldn't work!

Now, if the 50 lane motorway ever got full, then true enough there isn't any prioritisation before the prioritisation three lane motorway, but as has already been explained that isn't likely to happen due the nature of TCP/IP.

I'm not sure Dave would put his name to that, or if it's any help, but I thought it worth a try!

Ian

Edited by IanWild (Thu 10-Jan-08 18:20:24)

Really?

My connection on Sky is perfect for....

Gaming, P2P, and HTTP, VPN, RDP, and every thing else I have tried. As well as no limits or traffic shaping.

In reply to:

---

Because the 3 lane motorway is the bottleneck, there isn't room for all the traffic that wants to get through.

---

That's my point Ian, if hypothetically, the congestion occurs prior to reaching the PN network all traffic will be slower as it hasn't reached the point at which PN can mark it with a coloured blob yet.

I accept that applies to all ISP's. I now understand better how HTTP and games traverse PN's network faster than other protocols. I am trying to put the various claims into perspective.

To claim faster HTTP and gaming traffic than others without the caveat that this only applies on the host network is slightly misleading. Due to other circumstances outside of the particular networks control.

This is why I question what jelv wrote.