All of Worktree’s services are now hosted on our own hardware, with traffic flowing through our own network and IPs, colocated in a local Canadian ISP’s datacentre. Worktree has achieved 100% infrastructure sovereignty.

Our journey towards fully-sovereign infrastructure has been rocky. Being a fully bootstrapped company has meant we have had to compromise on provider choices to optimize for lower recurring costs over capital investment.

For the past year we’ve been on OVH Bare Metal servers. This has worked well: features like vRack and off-link IP pools bridged the gap between “rented servers” and “colocation” enough we could launch complex products like Worktree Cloud’s CDN and Actions 2.0, both of which have abnormally complex network requirements that make it difficult to build on someone else’s infrastructure.

But, ultimately, OVH is a French company. Arguably more tasteful than a big provider like Amazon or Microsoft, but still not the “fully Canadian” sovereignty we want. To achieve both the advanced networking and high compute we need, there’s really only one path forward: we build it ourselves.

Moving to Colocation

Over the past few months we have been buying server hardware, hard drives, SSDs, network equipment, fibre cables… Unfortunately we chose to do this during multiple worldwide catastrophies and economic bubbles, which has meant it’s been increasingly expensive and a slow process.

However, we eventually amassed enough hardware to replace everything that we were previously paying OVH to host.

We got in touch with a local fibre ISP, who just happened to have space available in a datacentre a few minutes away from our office here in Montréal. We couldn’t have asked for a better first location. Who wouldn’t want to be able to take the Métro to replace a hard drive?

So our physical sovereignty is solved; we have space in a datacentre with our own first-party servers. That brings us to the next challenge: the network.

AS402625

While we could just rent IP space from an upstream ISP, we really need to own our own network. By owning IP space and operating the core routers first-party, we can add resilience through multiple upstream internet peers and anycast routing between multiple points-of-presence. Also, especially once we start handing IPs out to Cloud customers, having long-term assignments is a huge plus; we have these same IPs forever, even if we change providers, move to different datacentres, etc.

So, in what has been the largest capital investment in Worktree to date, we now operate our own slice of the internet: AS402625, with a large assignment of IPv6, and a purchase of a small amount of IPv4. (As a fun historical note, the IPv4 block was purchased off of the Canada Post Corporation, which is an amusing bit of Canadian heritage buried in our network.)

All Worktree services have moved to these new IP blocks, announced from our colocation here in Montréal. With this, every layer from the packets to the hard drives is owned by us, and the datacentre it runs in is fully Canadian-owned.

This has been a long and complicated (and expensive…) journey, but the result was worth it. This new infrastructure and network lays the foundation for an incredible roadmap of ambitious and exciting products we’re already building for the next generation of Worktree and Worktree Cloud.

Watch this space…