Figure 1: Data hall schematic (TSM tutorials)
The data centre exists to house IT equipment. Broadly, this can be broken down into:
The data hall is the main area of interest in a data centre where IT equipment is housed, section 1. Data centre environments may have one or more data halls, sometimes called suites.
The data hall often has a raised floor, meaning that the tiles can be removed and a space exists under the floor, subsection 2. It is somewhat similar to a false-ceiling in reverse.
Not all data centres have raised floors. Where present, the raised floor has two key functions:
All data halls will vary. However, most will include some/all of the following:
A cage is a subdivision of a data hall to demarcate a number of racks for security or regulatory purposes.
IT equipment in data centre environments is normally rack-mount form. Rack-mount equipment has a standard width of 48.3 cm or 19 inches, which includes the rack ears protruding from the side of each piece of equipment.
Equipment height is standardised in multiples of 44.5 mm or 1.752 inches. One rack unit (commonly called 1U) is equivalent to 44.5 mm. Conventionally we just use the number of units, such as 4U.
You will sometimes encounter so-called Blade Servers in high-performance computing environments. A blade system consists of a chassis that holds a number of individual blade servers, subsection 4. Each blade server is a fully functional server but is powered and may receive connectivity via a backplane in the enclosure. Blade servers needs to be managed carefully as their power density is very high.
Non-rackmount equipment such as tower-type server PCs should be discouraged in data centre environments. Sometimes legacy equipment has to be accomodated. The only exception to this would be large equipment like mainframe computers which are supplied in a cabinet case, subsection 5.
IT equipment is mounted in racks within cabinets. Cabinets make up rows and pairs of rows make up aisles.
A cabinet contains front-and-back vertical rack rails, and is secured by a lockable door, usually perforated. Equipment is racked within the cabinet. Standard full-size cabinets are 42U in height, with each position numbered from the bottom upwards.
Most servers, like desktop computers, are air cooled. Air is drawn in at the front and ejected out the rear of the server. Servers are always racked front-to-back.
Cabling is normally done at the rear of the rack. Networking and power distribution equipment is often (but not always) mounted at the rear of a cabinet facing the rear.
Racks have holes drilled in them for allowing equipment to be secured. These are normally unthreaded, and allow cage nuts and bolts to be inserted. The cage nut inserts into the rack, whilst the bolt is inserted through the holes drilled in the IT equipment’s rack ears.
Heavier pieces of equipment such as servers normally are placed on horizontal rails inserted into the rack first. The rails are secured to the front and back racks.
Where more than one cabinet is required, they are normally placed beside each other to form a row. The number of cabinets in a row will depend on spatial and usage requirements.
An aisle is where two rows of cabinets face each other. Data centres normally use a hot aisle/cold aisle arrangement where servers face each other front-on or rear-on. This is to aid in cooling (see later).
There are a number of trends globally that are influencing data centre design and usage:
Many organisations operate some on-site data centre provision:
An on-site data centre will need WAN connections to be ordered. These may take a number of forms such as standard DSL, GPON fibre, leased lines, microwave links. Carrier-owned equipment is often required and connects to customer owned equipment at the so-called demarcation point.
An on-site data centre located within a larger facility will also often form the central hub of the end-user-facing networking throughout the building or site, such as the office LAN.
On-site data centres need careful co-ordination of different responsibilities:
Where exactly the demarcation between both sets of staff occurs is fluid and organisation-dependent. However, it is essential that the IT staff have a basic knowledge of certain mechanical and electrical concepts to facilitate productive interaction.
Co-location is where a customer rents space from a so-called co-location facility, or co-lo.
Co-los rent space in various different units:
Within a co-lo environment, the customer is solely responsible for their own networking. This means that the co-lo provider does not supply IP addresses or outbound connectivity as standard. Customers must have the appropriate networking cables and equipment to adequately connect their own equipment together.
Customers may sometimes have equipment in multiple cabinets, or may wish to connect directly to other data centre customers. They can do this by ordering a cross-connect from the co-lo provider. Depending on the provider, there may be a charge for this service.
Most links to locations away from the data centre will be made via telecommunication carriers.
Whilst some data centres historically have restricted the carriers a customer can use, indeed many have been operated by telecommunications carriers themselves, most are said to be carrier-neutral.
Carriers normally terminate their connections in a so-called meet-me room (MMR). The meet-me room allows patching through cross-connects to the customer’s equipment in the data hall.
It is the responsibility of the customer to order the WAN service from the carrier, and to order the appropriate cross connect from the co-lo provider.
Business needs must drive the choice of data centre provision. Some key questions include:
The solution chosen is often a compromise, and frequently a combination of multiple modes. For example:
