Sometimes the terminology used to describe datacentres seems just as complicated as the facilities themselves. Frustratingly, datacentres are very difficult to compare without delving into technical, jargon and acronym-heavy spec sheets. This guide aims to distill some of the main concepts that will help you to choose the best facility for your business.
UPS – an Uninterruptible Power Supply (UPS) is a large battery (or sometimes fly-wheel) power source that provides emergency electricity to the datacentre if mains power fails. It activates nearly instantaneously, and powers the facility for a limited amount of time until standby power sources (such as diesel generators) are activated.
Power density – the amount of electricity that can be used in each rack. This is usually determined both by the facility’s access to mains power grids, backup systems and its cooling technology. Most facilities can accommodate up to ~ 5 kVA without issues; higher densities may be possible with extra cooling, but multiple racks with a lower power density may be cheaper.
Standby power – an emergency source of power, usually generators, that will ensure uninterrupted power to all components inside a datacentre. Since generators can take a few minutes to start up and carry a full load, standby power sources are often bridged with UPS units (defined above) for uninterrupted power.
N+1 redundancy – parallel redundancy, with a N+1 system including a primary system in addition to a redundant system (+1). N+2 redundancy would indicate 2 independent backup systems, in addition to the primary system.
PDU – Power Distribution Units (PDUs) – unsurprisingly – distribute power to your equipment. They are best thought of as power boards that are mounted to the side of your rack. Rather than household outlets, PDUs usually have 20-24 IEC “kettle cord” style outlets.
Amps / kW / kVA – these are all measurements of power draw. Many facilities measure power usage in different scientific units, so it’s best to double check. Make sure you purchase and plan for sufficient power – it’s a good idea to have at least 25% of unallocated power to allow for growth and higher loads.
CRAC unit – Computer Room Air Conditioning (CRAC) units are cooling devices used to monitor and maintain temperature, humidity and air distribution in technical areas of a datacentre. CRAC units are a standard feature of modern datacentres, ensuring the efficient operation of servers and equipment.
IDEC unit – Indirect Evaporative Cooling provides low-humidity air-conditioning into a datacentre using a method that cools fresh outside air without adding any humidity before entering the building. IDEC is a more effective cooling system due to this process, and is also more energy-efficient. It is not (yet) common in Australian datacentres, but is used in Sydney’s Equinix SY4.
VESDA – Very Early Smoke Detection Apparatus works by continuously sampling air to detect any sign of smoke. This is based on laser based smoke detection and consists of a network of pipes that covers the datacentre ceiling.
Pre-action dry pipe sprinklers – a sprinkler system in which water is withheld from the pipes until an actual fire is detected. This means that pipes cannot leak in water-sensitive environments, keeping your equipment nice and dry (unless there really is a fire).
N+x% redundancy – similar to N+1, with a primary system (N) and an additional indicated number of redundant systems (+x).
Cross connect – a physicalcable – usually fibre – that connects one datacentre network to another. For instance, if you plan to operate your own routers and connect to several IP transit providers, you would need a cross connect for each. Cross connects are low-latency, dedicated point-to-point links within a facility or campus.
Internet exchange / peering – a network where telcos, ISPs and major companies exchange internet traffic. Internet exchanges are usually offer lower cost, and faster connectivity, but only to selected networks that have subscribed and connected to the exchange. For instance, Intergrid connects to NSW-IX in Sydney where more than 190 networks exchange traffic.
IP transit – best thought of as an internet service in the datacentre; like your internet at home, but delivered to your datacentre rack. There are differences, though; IP transit is often more expensive than a household internet service, but may have enterprise-grade uptime guarantees. It may also support BGP (Border Gateway Protocol) for route management.
Facility & services
Technical space – data halls, where racks are located. Technical space does not include foyers, meeting rooms, etcetera.
Private cage – A private, enclosed area within a datacentre’s colocation space, secured with mesh walls on all sides and a locking metal door as it’s point of entry. Only the cage owners have and can provide access to their own cage.
Biometric access – physical two-factor authentication for entry into secure areas of a facility. In datacentres, this usually involves scanning an access card against a RFID reader, and then electronic fingerprint verification. In addition to confirming fingerprints, these readers often also check the temperature of your finger… to make sure it’s alive.
Remote hands – staff available for performing physical tasks under specific instructions; for example, removing a disk or power cycling a server. Usually attracts ad-hoc fees, but is useful for remote sites or emergency situations.
Sundries – generic spare components – such as cage nuts or patch leads – that are available on-site for purchase. Vendor specific components are rarely available.