Open Compute Infrastructure has a rack scale design level, and this is with good reason. The holistic approach in designing Open Compute hardware has enabled the community to develop a rack that truly is groundbreaking in many ways.
The inner width dimension of an Open Rack is 2 inches wider than a standard 19” EIA server rack. Although this might not seem pivotal, the additional two inches make all the difference when it comes down to the compute density of the rack. While a conventional rack can hold 4 disks in the length, an Open Rack will be able to hold one additional disk within the 21” width. This translates to huge disk capacity gains on server and rack level. An open rack can have up to 30% more disk capacity compared to a 19” rack yet the outer dimension width is exactly the same.
Unlike conventional servers, Open Compute Servers get powered from the power bar(s) centrally from the back of the rack. This saves hundreds of unnecessary components in the server. A power shelf (and optional battery back up units) is placed in-rack, making investments in a central power room unnecessary. This obviously translates in huge cost savings when designing a datacenter. Not only does in-rack power architecture lead to higher operational flexibility as well as higher availability, it also provides investment flexibility (pay-as-you-grow) in stead of huge investments in a power distribution room, including UPS’s, switchgear transfer switches and a separate fire-rated battery room. We have not even covered the operational savings and valuable square meters that can be saved. The distributed in-rack power architecture also enables the use of less power conversions leading to significant power savings…
The opening in the front of the Open Rack is larger, which improves airflow and creates a good thermal profile. This allows for higher compute density. The rack depth makes serviceability more practical and thermal flows more efficient. The front end of the rack also provides clean cable management.