Rack and cable management is an incredibly underrated topic in most infrastructure projects. It is often a double-edged sword in that it is far easier and costs less money in time and materials to do it the wrong way. Convincing your customer, or your manager for that matter, that they should spend more resource in “making it look pretty” is a conversation that will more often than not result in you being told to “just plug it in”.
However, if you can get past the short-term costs, there is significant long-term value in making sure that your rack hardware and server room is set up properly from the very start of a project. The key to doing this efficiently and within budget is to make sure it is addressed at an architectural level, rather than at an engineering one. The difference between pre-planning rack layouts and cable runs before any hardware is ordered vs ordering what works best for the network architecture with no thought to where each component will eventually live, is the difference between an easy to manage server rack and a rats nest. Just as importantly as the cables at the front of the rack, are the power feeds and cables at the back, which is the target of this blog post.
For 20 years Italik have been designing networks from the ground up and building on each project with improvements for the next one. Wherever possible we always try to have distributed power to each rack in a server room. It is standard stuff, each rack has a minimum of two PDU’s, and each PDU is fed from a different mains input line via UPS. We always pre-plan the PDU output lines so we know which devices are being fed from which PDU so that we can ensure devices have redundant power links. We always label every PDU and power link so that power to each device can be easily traced back to the power source. In the event of a mains input failure, critical devices always have redundant power and can keep core services running.
Inevitably, no matter the design, 5 years into a live site, with real users and real engineers, some of the power leads have been swapped, some of the labels have been lost, devices have failed and been replaced or entirely removed. The more iterations of hardware in a rack, the more difficult it becomes for engineers to track the initial intentions of architected, distributed power.
This year, we made a very small change at one of our new installs which hopefully will have an impact in the day-to-day running of the site and may even help in the event of power loss or critical device failure. We followed the same architectural approach, but we made a “1st line” change by buying two different colours of power cables, rather than using the ones that come with the device. It sounds simple (because it is) but we believe the difference for our engineers will be significant. If a device has redundant power options, it gets a red and a black input. If it only has a single PSU, then the input should not be the same colour as its neighbour. It is the same solution, but with a much clearer “at a glance” impact, that may save the customer hours in the event of a critical failure.
To Follow – Front facing rack design and management