How to Make a Virtual Data Center with your Existing Infrastructure

Before starting Resilio, one of the coolest applications of BitTorrent technology I ever saw was when some of the world’s largest technology companies started using the protocol to upgrade operating systems at massive scale in minimal time.

I believe Twitter was the first, with a system written by Larry Gaeda (now the founder/CEO of Envoy, the maker of those iPads you sign into when visiting tech companies). His software reduced hours of update intervals using legacy client-server based tools to mere seconds instead, using peer to peer software over 10GBe backplanes in the data center.

The concept worked so well that, before long, Facebook, ServiceNow, Etsy and others followed suit and had similar tools in place.

It was a clever use of existing infrastructure to power a massive data distribution problem.

Get more from your existing infrastructure

Think about it. In the client-server model (the way their legacy patching system worked), you had to have enough server capacity to match or exceed client demand.

Flash upgrade operations of all servers meant that you had to double your server farm just to update your server farm (or rate limit the updates and leave critical security vulnerabilities unpatched).

Alternatively, in peer to peer models, every “client” is also a “server”, allowing very efficient distribution over high capacity LANs.

It’s a great way to get more out of infrastructure you already have, and this general method has more applicability than you think.

An idle machine is a resource

For instance, if you are reading this in a large technology office right now (and assuming you have an open floor plan), look out across the expanse of the office. You probably see table after table stacked with IT equipment.

And depending on the time of day or the culture, many of those workstations are likely empty and idle as meetings are underway, people are traveling or it’s after hours.

Every idle machine is effectively wasted capital, depreciating away.

What if you could harness all of those idle systems with a little piece of software and create a large virtual data center, just like Twitter and others did when they used servers to update other servers?  How much capacity would that be if you put it all together?

Resilio puts idle capacity to work

That’s what many of Resilio’s customers are able to do. With a Resilio agent on any machine, we can put that machine to work on hard tasks like scaling data distribution or remotely operating workflows.

A good example of this is the distribution of large software builds across a LAN to various test machines and QA professionals. Instead of building a data center in each office to serve the builds (which needs to be capable enough to match demand/or rate limited in some way), let every xBox and desktop distribute the build to neighbors on the LAN. Add what’s great about this method is the more machines you leverage, the faster the overall system gets.

This inherent speed in the system is why upgrading 10,000 servers happens in seconds. And Resilio can put your idle capacity to work as well, making your workflows faster.