In my article in the first and third issues of the InterGlobix Magazine, I described how the London Internet Exchange (LINX) came into existence in 1994, as a place for ISPs to exchange their traffic, and then how things developed from there.
I left the story in 2013, after we had navigated through the pressures of the London Olympic Games in 2012, and a very significant change in the main networking equipment that we used.
The story continues…
Regional Peering In The Uk
The UK is horribly London centric, and many people worry about this for all sorts of reasons.
The Thames flood barrier is a piece of critical infrastructure in the UK, since if the river Thames flooded central London (and especially the underground train network) recovery would probably would be very long and painful.
The London centric nature of the UK extends to networking, and I have long used this Roman map to illustrate the point.
It shows that the road network in the UK still follows the pathways set 2,000 years ago. This was copied by railways and canals, and later—unsurprisingly—by telecommunications networks.
The relevance of this is that if you were looking to build resilient interconnections points in the UK, where would you put them?
I published a strategy paper about this in early 2012, and it contained some fateful words.
The logic that brought LINX and many other IXPs into existence in the first place was to “keep traffic local”. Two notable ironies stem from these origins:
- LINX (and other similar IXPs) have become international peering points, with considerable traffic being exchanged that has nothing much to do with the geographical location of the IXP.
- LINX (and others like us) have become quite dominant in our domestic markets, and it seems as if we have not only inhibited the development of regional IXPs in our countries, but our existence and success has actively encouraged the backhaul of Internet traffic to our locations.
We can also note that:
- Several regions of the UK have populations (with corresponding high broadband Internet penetration and usage) that are larger than some EU states;
- LINX (and others like us) could probably succeed in establishing regional IXPs around the country in a way that previous one-off initiatives (e.g. in Scotland) have not managed to do;
- We understand much more clearly now (than perhaps we did in the pioneering days) that IXPs are places where access (ISPs) meets content, and they can work with the various parties to engineer this to happen;
- We can observe successful regional IXP approaches and models in other countries, and use this to generate a successful model for the UK;
- More local/regional peering could reduce some of the pressure of extreme growth on the core infrastructure of LINX in London;
- The general increase in traffic everywhere should mean that regional exchange points can reach critical mass more quickly than would have been true in the past.
[Note: All of those words are straight from the original strategy document.]
So, we have the irony that LINX had probably, quite unintentionally, inhibited the growth of regional peering in the UK. And as the famous saying goes, if you are part of the problem, then perhaps you need to think about being part of the solution!
Many thought that it was too late to change the way that interconnection had developed in the UK, but LINX exists for the benefit of its members, so it fell to us to try to do something about it.
We established a few principles to guide us in this endeavour:
- They are only to be built where there is clear network operator support for it;
- They are targeted at places where there is a good prospect of them achieving viability;
- Ideally, they need to be hosted in carrier neutral data centres, so that members have access under agreeable terms.
These were not easy principles to satisfy, and I could probably write a book never mind a magazine article, on the struggle that we went through in the years that followed.
Despite it being London Olympic year, LINX’s Manchester members asked us if we would build a new Internet exchange there—and do it the same year! Looking back, it was a pretty crazy thing to try, but we managed it. The key advantage of Manchester is that it had a well developed data centre scene, much more so than any other UK city outside of London.
It is worth noting briefly that there had previously been an IXP in Manchester (MANAP), but that it had failed, and ceased to function. This is a rather sad reflection on how hard it is to establish an exchange, and reinforcement for the idea that a more robust organization (such as LINX) is somewhat necessary when you are trying to do pioneering work before the economics catches up.