Author: Mike Hughes

Just last month, in mid-September, Andy Davidson brought up the switch at IXLeeds, the latest UK regional IXP.

You’ll note I say “the latest”, but how many non-London UK IXPs can you name off the top of your head? Not many, I’ll wager. Fewer that are still operating, too. No, the LINX PoP in Slough doesn’t count in my picture of non-London!

This is the problem: It’s often said that there isn’t the level of regional IP peering going on in the UK that there probably should be for redundancy reasons. The majority of IP peering in the UK happens in London, and when it isn’t happening in London, it’s probably happening in Amsterdam instead.

Let’s face it, on an island that’s ~15ms round-trip top-to-bottom, we’re less likely to peer to reduce latency, especially when the architecture of the incumbent wholesale DSL platform doesn’t encourage networks to do little beside haul all broadband customer traffic to a central point before dispersing it.

Previous attempts at establishing regional IXPs in the UK have had varying levels of success. The most successful to date, in terms of number of participants and achieving critical mass is probably MANAP – which was founded in Manchester in 1997.

Unlike LINX which did survive (well, successfully resisted) a demutualisation attempt, MANAP only sort-of did. It allowed it’s infrastructure to be taken over by a company funded by the local Regional Development Association, and the exchange became a service provided over the infrastructure which was no longer dedicated to IXP operations, but also carried other traffic and provided other services.

The MANAP that exists today is not the same exchange, it has been subsumed into the NWIX platform and operates as Edge-IX, a distributed exchange which is present in both Manchester, elsewhere in the Northwest, and in many other locations, including those in London’s Docklands that it was initially intended to provide redundancy for. It’s has a different flavour, and has lost some elements of it’s “regionality”.

What distinguishes it from a carrier, other than the Edge-IX services being non-profit, while the NWIX ones are?

I’m not suggesting that this is a better or worse model, just different, and probably not regional anymore. If this, i.e. reinventing yourself as an inter-regional IX, is the only way a regional IXP in the UK can survive, then we’ll find it very challenging to reach position of sustainable regional peering in the UK. Could things have been different in Manchester?

You may be questioning what issue I have with a “wide-area” exchange point, distributed over a large geographic area? The main concern is shared fate. A disruption that would otherwise be localised, spreading easily. I can probably write a whole article on that. Maybe I will another day…

So, why would a quick hop over the Pennines to Leeds be any different?

Manchester itself is also at risk of being unattractive as a location for regional IXs – with the recent purchase of IFL by Telecity Group there is very little organisational diversity or competition in the Manchester co-lo market – there’s existing facilities such as Vialtus Serverbank, and recent new entrant Ice Colo. Folks in the Manchester area were very quick on the social networks to state their fears about anticipated price rises and few options as a result of the lack of choice.

The Leeds scene is rather different, with lots of smaller, entrepreneurial companies active in the metro area. This is a double-edged sword, as while it results in competition in the co-lo market which folks like, it also meant that IXLeeds couldn’t be present everywhere the potential IX participants wanted to connect, certainly from day one. There’s a future aspiration to expand within the metro.

One of the early strengths in IXLeeds is that has a good community feel behind it, including the involvement of folks who have experienced peering in Manchester, while the Yorkshire RDA have been involved from the outset in getting folks together, but (so far) haven’t felt the temptation to get in the driving seat, instead choosing to play the role of facilitator.

There’s a will to succeed, so hopefully they will reach the critical mass that is required to sustain the exchange.

A concern I have is the lack of international capacity into the Leeds area, Manchester is in a better position here due to the independant (from London) Transatlantic connectivity arriving in the area.

That said, while international bandwidth a something of a pre-requisite for a national exchange point, is that actually necessary for a successful regional IX?

Then again – what are the success criteria for an IX, especially a regional one? A graph that forever goes up and to the right? They probably are and should be different from a national IX. Is the regional IX not being satisfied with it’s lot, and wanting to be like it’s larger neighbour, what actually destroys it? Maybe that’s another article in itself?

I’d say it depends on how non-London-centric the early IXLeeds ISPs are, how much of their traffic is delivered locally, and how much traffic they have between each other that they might normally route through London.

If my previous experience is anything to go by – such as opening a new PoP for an already successful IX – these things usually “slow start” – so that means patience is required.

I’m going to come back to this topic in the coming weeks, I’ll try and write about some of the side issues I’ve threatened to cover above, and maybe touch on a missed opportunity.

Still think they should have called it the Rhubarb Internet Exchange. Even if it was just to confuse people. 🙂