Deterministic gateway selection

I noticed buried away in the GEA migration thread this little nugget from hazey_flakes that "soon" gateway selection would be more deterministically assigned based on location.

In reply to a post by hazey_flakes:

At present the gateway is purely based on whichever gateway responds fastest to the pppoe initialisation request from the CPE. We are making a change soon which will make this more deterministic and you'll generally (non fault conditions) be connected to a BNG pool closer to your location.

Now this is something I care a lot about and try to keep an eye on any developments, but this is the first I've seen of this, I wondered if anybody knows of anything more about what's going on in the public domain, especially timescales? (Or if Brandon can be persuaded into telling us more!)

I'd also be interested to know why latency is doubled for most things when I'm on a Manchester gateway. I'm not far from Manchester so you'd think it would make more sense for me to use a Manchester gateway, but it's as if my exchange is connected to Zen's network via London instead for some reason.

It's all about distance travelled.

Connecting to Manchester if you are closer there than London only benefits you if the servers you are connecting to are also near Manchester, and there is a route from Zen's gateway in Manchester to those local servers. However the vast majority of servers and routings go via London, so you don't gain much really. You may be worse of if the distance travelled from your home to Manchester then down to London is greater than if you simply went to London directly, and that is what Zen are often criticised for allowing to happening. There have been cases of someone that should have < 2ms pings to London servers as they are London, but seeing pings significantly higher because they go from London up to Manchester then back down to London.

I have often complained about this issue, but for the past two months I've only been connected to the lowest latency (closest) gateways. That's the longest ever stretch of time I've seen.

Which is superb!

I don't know if that means it's already live, or is being selectively tested (and I'm an unknowing but grateful test subject), but it would appear something good is happening behind the scenes.

This doesn't seem to fit what I see.

I'm to the west of Manchester. The distance from here to London and from here to Manchester then London are really not very different at all.

If I'm connected to a Manchester gateway, the latency to the gateway itself is noticeably higher (not just a few ms) than if I'm on a London gateway. How does that make any sense when Manchester is less than 50 miles away and London is more than 200?

FakeJake

It makes no sense to me as your Physical routing may well go via Manchester to get to a London gateway!

It would only make sense if you are on a physical routing that bypasses Manchester going direct to London then back to the Manchester Gateway. This may happen in Cheshire or around Stoke may even happen n Liverpool if they have built a ring (Eg London, Sheffield,, Leeds, MR, Liverpool, Chester, Birmingham, London) Go the wrong way around the ring can then add ms to what you would think is a shorter hop!

I guess so. I'm on the outskirts of Chester so perhaps I really am connected to London instead.

It's a shame that when using my work's VPN, my traffic goes to London to come back up to a data centre that's less than 3 miles away. Perhaps I should set up a wireless link instead and ditch the VPN 😉

In reply to a post by FakeJake:

I guess so. I'm on the outskirts of Chester so perhaps I really am connected to London instead.

It's a shame that when using my work's VPN, my traffic goes to London to come back up to a data centre that's less than 3 miles away. Perhaps I should set up a wireless link instead and ditch the VPN 😉

This might be dependent on whom they are using for the backhaul as I don't think they have their own dark fibre (though wouldn't be surpised if they intend to at some point in the future)

Maybe worthwhile having a chat with technical support if it's causing you a problem for work?

Honestly it's all first world problems for me. I can just use RDP and keep latency and bandwidth sensitive activity within the workplace network.

Plus I can drive/cycle/walk to said data centre if I need to.

Having less latency would be nice but not needed.

FakeJake

Many years back we used to put big files on a USB ( Or CD-Rom) and do just that, physically take the data from one building to another. We used to joke that for slightly further a homing pigeon would be the fastest as it misses the traffic jams