Building a new data center, or expanding an existing one, is a monumental task. When it’s done right, our clients don’t notice a thing, other than the influx of available dedicated servers. But behind the scenes, it’s a real labor of love, with a lot of excitement and a lot of hard work.
As SingleHop’s head of data center operations, one of my biggest responsibilities is managing data center expansions and ensuring seamless integration of new servers into our existing infrastructure. Our new latest expansion into CR-10 is the third data center I have helped set up, and the second one I have set up from scratch. In this post and the next one, I’d like to give you a small glimpse into what goes on when we add new data center space.
Part I: Core Cabinets and Support
After we’ve chosen our new data center location, the first step is to set up “core cabinets” in the space. These cabinets house our primary networking gear, internal servers, and other support devices for the space. The core cabinets are installed in a central location, and each one holds two power distribution units on separate 30-amp circuits.
Here at SingleHop, we turn up two core cabinets for every 80 client cabinets. Each client cabinet has half of its networking connected to each of the two core cabinets. A client cabinet’s primary switch is fed by both routers in each core cabinet, ensuring redundancy in the event of a core switch failure. These switches are racked by myself and configured by our Senior Network Administrator, and then linked to the outside network.
Once the core switches are in place and configured, we can get our other internal devices online. Shared firewall solutions, shared load balancers, our provisioning system, various internal servers, and a myriad of other devices populate these cabinets.
After we install all of the necessary equipment in the cores, we punch down and link four cross connects from each client cabinet to each core. As you might expect, this makes for some very dense wiring situations. With every new data center we turn up, we continue to optimize our cable management methods to ensure optimal cooling and organization. As a result, any cable in the data center can be traced from one end to the other with minimal effort, and all of our core equipment sits at a nominal temperature with ample airflow. It might seem like a small task if your putting a home network together, but when you are managing miles and miles of critical network cables, perfection is the only option.