Overview of network changes

Back to the overview
Overview of network changes
4 minute read

In the past 6 months we did a major network overhaul. In this post we’ll outline a couple of the improvements we made and the things we’ve done to enable future growth of the Tilaa network.

Core network

We did a core network upgrade, allowing us to connect more racks and servers to our network while increasing bandwidth. Each of our server racks now has multiple 10GbE connections to our core network.

We’ve migrated from 3rd party MPLS links to our own dark fiber for inter-DC connectivity, provided by Relined. This increases our network independence and allows us to easily increase inter-DC bandwidth or provision independent network paths by using WDM (Wavelength-division multiplexing) equipment. We have multiple 10GbE inter-DC connections in use which can be further extended in the future when required.

Internet connectivity

We’ve added a new Tier 1 provider for transit connectivity: Tata communications (AS6453). This is an additional to our Tier 1 transit connectivity from NTT (AS2914). Together these two transit connections provide us with superb connectivity to practically all corners of the earth (well, the ones that matter ;).

We’ve upgraded our peering connections to the AMS-IX and NL-IX internet exchanges to 10GbE and have added a lot of direct peering sessions with other networks over the last months (and will of course continue to add more in the future).

We’ve also upgraded our Openpeering partial transit package from NL-transit to EU-transit, which further improves our network paths and reduces network latency to lots of networks we don’t have direct peering sessions with through other internet exchanges (such as DE-CIX and LINX) and private interconnects (UPC, KPN, etc).

All our links to the internet are now at least 10GbE.

Additional prefix (/19)

Some of you may be aware that the RIPE region has run out of IPv4 space. If not, I highly recommend you read the Wikipedia article on the topic of IPv4 address exhaustion. What it boils down to is that networks are unable to obtain new IPv4 addresses. This is not currently an imminent problem for a lot of providers, because they often either have a large pool of IPv4 space available or they just don’t grow that much anymore. Unfortunately we don’t have that luxury: Our current remaining pool of IPv4 addresses is shrinking and at our current rate of growth we’ll run out of them somewhere next year.

The IPv6 protocol that is designed to replace IPv4 has been fully deployed by only a small fraction of ISP’s (of which we are one), which means that providing IPv4 connectivity to our clients will be required for the foreseeable future until at least 90% of internet users have access to IPv6 connectivity. We estimate that it will take at least 3 to 5 years to get there. It would definitely help if the large access providers (in the case of the Netherlands: UPC, Ziggo, KPN) would deploy IPv6 to their end customers. If they take the lead I’m sure others will quickly follow.

Until then we have to ensure we have enough IPv4 space to continue to grow after 2014. Therefore we’ve bought /19 IPv4 space from another company, which holds 8192 IP addresses. Whether or not this is enough to get to the point where IPv6 connectivity is common remains to be seen.

Share this article
Back to the overview
More like this