If networks could talk

Get an insight into the latest developments and key considerations that go into network design and development here at NL-ix with Dirk Kalkman, NL-ix Network Architect.

I began working at NL-ix on 1st January 2020, so I am now looking back on my first year with the company. It has been an extraordinarily busy year packed with lots of very satisfying projects, but it was also (as everyone knows) a very strange one.

Before I took up my new role I was Operational Manager for Spotler doing pretty much everything technical. From being a jack of all trades, I had to focus in on one seriously large network here at NL-ix. Luckily, the NL-ix Field Engineering team – Gert-Jan, Onno, Arno, Rob, and Timo (on the procurement side) – are amazing. They have endless expertise and historical knowledge between them so I can go to them with any question – optical issues, MPLS, VXS, whatever. Also, I was lucky to have a few months of getting to know the network first-hand before Covid put that on hold.

The Changing Network

The NL-ix network has improved a lot in 2020. New ultra-low-latency routes had already been negotiated from London across the channel to Amsterdam and Frankfurt before I joined, and these were hooked up, speeding up traffic and improving cross-channel diversity. The new routes are known as ‘trader routes’ due to their popularity with the high-speed trading community. Round Trip Delay (RTD) on the new routes is under 8.6 milliseconds from London to Frankfurt and 6 milliseconds from London to Amsterdam. The next big capacity design challenge we faced was in Frankfurt, which had been growing extremely fast.

We reviewed the capacity and routing, asking “what can we do to be more efficient?” And as we tackled the Frankfurt capacity issue we found the answer was “by being more ambitious,” The scope of the review soon grew to include Denmark, then Dusseldorf. The project snowballed, as there were so many efficiencies to be gained by tackling multiple routes at the same time.

A Long List of Lower Latencies

Working closely with procurement and leveraging the fact that NL-ix is not tied to any one provider, we were able to renegotiate, redesign and connect new lower-latency routes while keeping overheads low enough to avoid price increases.

We increased our three existing connections across Germany and Denmark to eleven and added 600 Gigabits of capacity to the existing backbone. The latency impacts were huge: we drove down the Amsterdam- Frankfurt primary latency to 5.8 ms with a 7.1 ms backup. In Copenhagen we reduced latency to Amsterdam by about a third (to 9 ms) by becoming one of the first networks to use the, now operational, new COBRAcable. Most other networks still run through a bottleneck at the Hamburg/Groningen handover. We also added a third separate backup route with a 10.2 ms latency. The Dusseldorf-Amsterdam connection also got better, with a primary link at just 3 ms and a backup link of 3.7 ms each way to Amsterdam and Frankfurt.

Every Millisecond Counts

It’s extremely satisfying to oversee such widespread increases in efficiency. There will be a huge impact on the speed of application delivery for our customers. As businesses move to distributed cloud applications, it’s not the thickness of the pipes that counts any more, it’s the efficiency of delivery.

We try to be proactive in our capacity and efficiency management - if we take a smart regular look at our network and constantly reroute traffic to knock off a few milliseconds, after a year or two we will have what is indisputably the fastest network in Europe. In fact we think we already have it, it’s just that nobody else dares to share their data, so it’s hard to be sure!

Next Up So, it’s been a busy year but there are many more challenges to come; France, Luxembourg and Belgium are in our sights for next year, as well as improvements at our network core here in the Netherlands, so watch this space!