January was a busy month for us

4 Feb 2021 | General, News, Development

January was a busy month for us

A busy month, January. Where it usually means a quiet start to the year, we had to work hard at Tuxis to move our stuff in Arnhem to Amsterdam.

Moving ?!

In addition to our locations in Ede (BIT-1, BIT-2A and BIT-2C), we have had a fourth data center where we have equipment running for years. Especially backup equipment and infrastructure that we want to keep online in the unlikely event that Ede would go offline.

In principle, that was Solcon and in 2018 it moved to Arnhem. It wasn’t very much then, so that was pretty easy. But in the past two years we have worked hard to grow dadup (to now more than 500TB of storage) and we have also connected our network to the internet via Arnhem.

We made the decision at the end of 2020; we are going to move our stuff in Arnhem to a datacenter in Amsterdam, maincubes. Thanks to ISPConnect (our industry association) we came into contact with maincubes, a small player in the Netherlands with a large parent organization in Germany. Privacy and collaboration are things that maincubes attaches great importance to, something that we at Tuxis also love.

Step 1
Connectivity

Choosing a new data center is one thing, making sure you have sufficient network capacity to it is also useful.

We contacted i4networks , who is in the process of rolling out a new service; DC2DC. This makes it relatively easy for them to provide redundant connections between a lot of different data centers in the Netherlands. At BIT they were already connected and at our request they also built a POP in maincubes AMS1.

We were live in early January. We had 10Gbit between Ede and Amsterdam. In addition to a POP at maincubes, we also placed equipment directly in Nikhef so that we can easily set up private interconnects with others.

Step 2
Equipment

The new locations must of course also be connected with physical equipment. In recent years, we have used Huawei switches in Ede with satisfaction.

Also now we have chosen Huawei, so that we have sufficient capacity. The switches have sufficient 10Gbit ports and also provide us with 40 / 100Gbit ports so that we can upgrade the connections from 10 to 40 or perhaps 100 Gigabit in the near future without making new investments.

With MPLS we can also easily provide our customers with connections between the different locations and easily connect new data centers.

Step 3
Move ahead

Well. How do you move 500TB of data to a data center 100KM away, without downtime? Thanks to Ceph, this was actually very simple (but not very efficient). We configured the VLANs in Arnhem along the entire path to Amsterdam, with which the two locations were actually linked together on layer 2. Then it is a matter of patience ..

Switch off a server in Arnhem, unscrew it from the rack, drive to Amsterdam, screw it into the rack, switch it on and wait a few minutes.

Ceph restored the missed changes to the cluster for the node that had moved within minutes; and no one who noticed.

And now? What’s the point of it?

Good question! You would say a data center is a data center. But that’s not entirely true.

Thanks to the cooperation with maincubes, we can also relatively easily provide services from Frankfurt, or to Frankfurt. And by setting up a network across Ede and two locations in Amsterdam, we can also set up connections with other parties more easily.

For example, since last week we have been active again on SPEED-IX, which we can connect ourselves to thanks to our presence at Nikhef. And next month we hope to move our AMS-IX connection in Ede to Amsterdam. This way we can minimize latency and complexity in our network, which in turn benefits reliability and stability.

And so our network is considerably expanded and even more flexible. The image gives a good impression.

daDup and Proxmox Backup Server

These are daDup and Proxmox Backup Server two services that we will serve from maincubes. These are services that require a lot of bandwidth, so it is useful to be, as many people call them, close to the backbone of the Internet .

And because backups from our clusters also go to those two services, we automatically separate geographically separated backups!