Last winter I wrote an enterprise storage dictionary for non-experts, and now it’s time for a data center version. The following are a dozen of the less-understood terms you’ll hear in the field. Post in the comments if you have suggestions for terms that should be on this list.
ASHRAE is the American Society of Heating, Refrigerating and Air-Conditioning Engineers. It’s a trade group, but the relevant parts are Technical Committee 1.5 (Computer Applications) and Technical Committee 9.9 (Mission Critical Facilities, Data Centers, Technology Spaces and Electronic Equipment).
Blades are modular stripped-down servers designed for space consolidation and quick maintenance. In a conventional rack, hundreds of blades can fit in the space of dozens of standard servers.
Close-coupled cooling is installed immediately next to a rack. This is useful for a rack with components where your standard data center cooling is insufficient.
Clusters are networked groups of servers. Modern clusters can be remotely dispersed, resulting in a virtual or software-defined cluster. Regardless of configuration, clusters are useful for applications that require multiple servers or storage working in tandem.
Containers are walled-off sections of your operating system reserved for specific applications. Containers are often used in virtual operating systems. The applications tend to be stripped-down in the form of microservices. Docker, RKT, and Mesos are popular container types. Kubernetes is an open-source program for container management. (Learn more in TechRepublic’s containers cheat sheet.)
DR is disaster recovery, which (to start) involves off-site data replication. The idea is to get your data center up and running as quickly as possible, even if it’s in a temporary or minimal configuration, while your real data center is repaired or rebuilt following the disaster. Many service providers offer DR, and just like all services you’ll want to test it now and then.
Hybrid clouds are systems where one or more public cloud services are combined with your own data center resources into a single virtually managed environment. That sounds great on paper, but the management can be incredibly complicated. Be skeptical if your vendor or systems integrator advertises a “hybrid cloud solution” along with words such as “easy” or “simple.” (Learn more in TechRepublic’s hybrid cloud cheat sheet.)
Hyperconverged is a term in flux: Ask 10 people what it means, and you may get 20 answers. I asked industry analyst Zeus Kerravala last year, who explained: “There was already a converged infrastructure market [lacking the software aspect] when this technology came around. Hyperconverged platforms are turnkey products that include all the hardware and software one needs to run a contained little data center in a box.” (Learn more in TechRepublic’s hyperconverged infrastructures cheat sheet.)
PDU is a power distribution unit. Sometimes this term is used in reference to your data center’s electrical system, but more commonly it means the power system inside a rack. PDUs can be tall power strips standing vertical up a rack post, or they can be centralized in conventional rackmount fashion. Sometimes these are combined with uninterruptable power supplies.
PUE is power usage effectiveness. It’s an industry standard (ISO/IEC 30134-2:2016, if you’re into the arcane) that measures the ratio of energy used to energy available. Perfect PUE would be 1. Most modern data centers have PUE of about 1.2-1.4. Anything above 1.5 means you should be looking for energy improvements. (Read how to save energy in your data center.)
Software-defined is the art of separating your network from your hardware. You can use virtualization software to create, change, scale, or delete software-defined networks at will. The networks can be dispersed or hybrids. You can even allow applications to do this for you, based on their own needs and policies that you set. That’s pretty cool. (Learn more in TechRepublic’s cheat sheet about software-defined data centers.)
Tiers, particularly as defined by the Uptime Institute, certify four levels of data centers. From the Uptime Institute’s website: Tier I is the most basic classification, and Tier IV is the most fault-tolerant. A data center provider called Switch defined its own Tier 5 level in 2017.