Live data from Hacker News

So you want to build your own data center

blog.railway.com

51–60 of 287 posts

Re: So you want to build your own data center

#51

Why would you call colocation "building your own data center"? You could call it "colocation" or "renting space in a data center". What are you building? You're racking. Can you say what you mean?

It seems a bit disingenuous but it’s common practice. Even the hyperscalers, who do have their own datacenters, include their colocation servers in the term “datacenter.” Good luck finding the actual, physical location of a server in GCP europe-west2-a (“London”). Maybe it’s in a real Google datacenter in London! Or it could be in an Equinix datacenter in Slough, one room away from AWS eu-west-1.

Cloudflare has also historically used “datacenter” to refer to their rack deployments.

All that said, for the purpose of the blog post, “building your own datacenter” is misleading.

Re: So you want to build your own data center

#52

Why would you call colocation "building your own data center"? You could call it "colocation" or "renting space in a data center". What are you building? You're racking. Can you say what you mean?

I have to second this. While it takes mich effort and in-depth knowledge do build up from an “empty” cage it’s still far from dealing with everything from building permits, to plan and realize a data center to code including redundant power lines, AC and fibre. Still kudos going this path in the cloud-centric time we live in.

Yes, the second is much more work, orders of magnitude at least.

Re: So you want to build your own data center

#53

Why would you call colocation "building your own data center"? You could call it "colocation" or "renting space in a data center". What are you building? You're racking. Can you say what you mean?

It seems a bit disingenuous but it’s common practice. Even the hyperscalers, who do have their own datacenters, include their colocation servers in the term “datacenter.” Good luck finding the actual, physical location of a server in GCP europe-west2-a (“London”). Maybe it’s in a real Google datacenter in London! Or it could be in an Equinix datacenter in Slough, one room away from AWS eu-west-1. Cloudflare has also…

The hyperscalers are absolutely not colo-ing their general purpose compute at Equinix! A cage for routers and direct connect, maybe some limited Edge CDN/compute at most.

Even where they do lease wholesale space, you'd be hard pushed to find examples of more than one in a single building. If you count them as Microsoft, Google, AWS then I'm not sure I can think of a single example off the top of my head. Only really possible if you start including players like IBM or Oracle in that list.

Re: So you want to build your own data center

#54
post #36

Earlier quoted context omitted.

I have to second this. While it takes mich effort and in-depth knowledge do build up from an “empty” cage it’s still far from dealing with everything from building permits, to plan and realize a data center to code including redundant power lines, AC and fibre. Still kudos going this path in the cloud-centric time we live in.

Having been around and through both, setting up a cage or two is very different than the entire facility.

I think you and GP are in agreement.

Re: So you want to build your own data center

#55
This is how you build a dominant company. Good for you ignoring the whiny conventional wisdom that keeps people stuck in the hyperscalers.

You’re an infrastructure company. You gotta own the metal that you sell or you’re just a middleman for the cloud, and always at risk of being undercut by a competitor on bare metal with $0 egress fees.

Colocation and peering for $0 egress is why Cloudflare has a free tier, and why new entrants could never compete with them by reselling cloud services.

In fact, for hyperscalers, bandwidth price gouging isn’t just a profit center; it’s a moat. It ensures you can’t build the next AWS on AWS, and creates an entirely new (and strategically weaker) market segment of “PaaS” on top of “IaaS.”

Re: So you want to build your own data center

#56
post #53

Earlier quoted context omitted.

It seems a bit disingenuous but it’s common practice. Even the hyperscalers, who do have their own datacenters, include their colocation servers in the term “datacenter.” Good luck finding the actual, physical location of a server in GCP europe-west2-a (“London”). Maybe it’s in a real Google datacenter in London! Or it could be in an Equinix datacenter in Slough, one room away from AWS eu-west-1. Cloudflare has also…

The hyperscalers are absolutely not colo-ing their general purpose compute at Equinix! A cage for routers and direct connect, maybe some limited Edge CDN/compute at most. Even where they do lease wholesale space, you'd be hard pushed to find examples of more than one in a single building. If you count them as Microsoft, Google, AWS then I'm not sure I can think of a single example off the top of my head. Only really…

Maybe leasing wholesale space shouldn’t be considered colocation, but GCP absolutely does this and the Slough datacenter was a real example.

I can’t dig up the source atm but IIRC some Equinix website was bragging about it (and it wasn’t just about direct connect to GCP).

Re: So you want to build your own data center

#57
post #18
post #15

Can anyone recommend some engineering reading for building and running DC infrastructure?

We didn't find many good up-to-date resources online on the hardware side of things - kinda why we wanted to write about it. The networking aspect was the most mystical - I highly recommend "BGP in the datacenter" by Dinesh Dutt on that (I think it's available for free via NVidia). Our design is heavily influenced by the ideas discussed there.

What was the background of your team going into this project? Did you hire specialists for it (whether full time or consultants)?

Re: So you want to build your own data center

#58
post #30

This is a pretty decent write up. One thing that comes to mind is why would you write your own internal tooling for managing a rack when Netbox exists? Netbox is fantastic and I wish I had this back in the mid 2000s when I was managing 50+ racks. https://github.com/netbox-community/netbox

It is not that difficult to build it into your app, if you're already storing information about hosts, networking etc. All you're really doing is expanding the scope, netbox is a fine starting point if you're willing to start there and build your systems around it, but if you've already got a system (or you need to do anything that doesn't fit netbox logic) you're probably better off just extending it.

In this case railway will need to care about a lot of extra information beyond just racks, IP addresses and physical servers.

Re: So you want to build your own data center

#59
post #53

Earlier quoted context omitted.

It seems a bit disingenuous but it’s common practice. Even the hyperscalers, who do have their own datacenters, include their colocation servers in the term “datacenter.” Good luck finding the actual, physical location of a server in GCP europe-west2-a (“London”). Maybe it’s in a real Google datacenter in London! Or it could be in an Equinix datacenter in Slough, one room away from AWS eu-west-1. Cloudflare has also…

The hyperscalers are absolutely not colo-ing their general purpose compute at Equinix! A cage for routers and direct connect, maybe some limited Edge CDN/compute at most. Even where they do lease wholesale space, you'd be hard pushed to find examples of more than one in a single building. If you count them as Microsoft, Google, AWS then I'm not sure I can think of a single example off the top of my head. Only really…

Hyperscalers use colos all the time for edge presence.

Re: So you want to build your own data center

#60

The date and time durations given seem a bit confusing to me... "we kicked off a Railway Metal project last year. Nine months later we were live with the first site in California". seems inconsistent with: "From kicking off the Railway Metal project in October last-year, it took us five long months to get the first servers plugged in" The article was posted today (Jan 2025), was it maybe originally written last year…

ah that's my bad - I wrote this in Dec, we only published in Jan. Obv. missed updating that.

Timeline wise; - we decided to go for it and spend the $$$ in Oct '23 - Convos/planning started ~ Jan '24 - Picked the vendors we wanted by ~ Feb/Mar '24 - Lead-times, etc... meant everything was ready for us to go fit the first gear by mostly ourselves at the start of May (that's the 5mo) - We did the "proper" re-install around June, followed closely by the second site in ~ Sep, around when we started letting our users on it as a open beta - Sep-Dec we just doubled down on refining software/automation and process while building out successive installs

Lead times can be mind numbing. We have certain switches from Arista that have a 3-6 mo leadtime. Servers are build to order, so again 2+ months depending on stock. And obv. holidays mean a lot of stuff shuts down around December.

Sometimes you can swap stuff around to get better lead-times, but then the operational complexity explodes because you have this slightly different component at this one site.

I used to be a EEE, and I thought supply chain there was bad. But with DCs I think it's sometimes worse because you don't directly control some parts of your BoM/supply chain (especially with build-to-order servers).

Post reply on HN