How do rack servers work?

Data centres consider rack units, IT cabinets and rack servers a first choice. Different designs and equipment of rack servers provide strong, demand-oriented computing power within a small space in a stable and durable rack enclosure.

Just three years after the in­tro­duc­tion of the first web server in 1990, Compaq’s ProLiant series was the first rack server to be developed. Ever since, rack servers have con­trib­uted to efficient computing and in­form­a­tion pro­cessing in data centres. They ensure that secure, cabinet-like en­clos­ures house critical servers and server com­pon­ents in a small space. Their in­tro­duc­tion was ac­com­pan­ied by the pro­lif­er­a­tion of the, now ubi­quit­ous, server rooms.

The term rack server describes the structure of the hardware and software ad­vant­ages. Rack servers consist of servers and com­pon­ents mounted in stand­ard­ised 19-inch racks as slide-in units. The rack housing includes mounting slots (rack bays) which mean hardware com­pon­ents can be screwed together securely and stably. The stacking principle ensures easy vent­il­a­tion of the server system and a higher density of hardware in a smaller space.

The enclosure and mounting grid known as a rack serves as a frame for housing the servers. In the rack, server hardware with server com­pon­ents can be mounted as ho­ri­zont­al or even vertical racks. With their stand­ard­ised width of 19 inches and various height units, rack servers can be ef­fi­ciently installed side by side. The in­stall­a­tion depth is 60, 80, 100 or 120 cen­ti­metres. Other important com­pon­ents are the guides, rails and brackets for cables, sensors, vent­il­a­tion and cooling units. Depending on safety re­quire­ments, they can also include systems for ex­tin­guish­ing gas in an emergency.

Other important com­pon­ents of rack servers are:

• Power supply (PSU)
• CPUs
• Random Access Memory (RAM)
• SSD or HDD
• Network cards or adapters
• RAID con­trol­ler
• Host bus adapter

Ap­plic­a­tions for rack servers are manifold. Due to their con­fig­ur­ab­il­ity and the simple scaling and expansion of server com­pon­ents, they’re suitable for small and medium-sized en­ter­prises as well as large data centers. By stacking servers in the rack, resources are con­sol­id­ated more ef­fi­ciently and in a space-saving manner.

The main tasks rack servers usually perform include:

• Vir­tu­al­isa­tion: Rack servers are suitable for vir­tu­al­iz­a­tion of hardware resources according to the XaaS (Anything-as-a-Service) principle because of their versatile con­fig­ur­a­tion and space-saving in­stall­a­tion. This enables IT in­fra­struc­tures to be ef­fect­ively reduced while main­tain­ing or improving per­form­ance.
• Com­mu­nic­a­tion: Rack servers control and manage internal and external com­mu­nic­a­tion processes.
• File server: Rack servers often serve as fileserv­ers. They store data and organise access to storage resources in the network.
• Print server: Server racks are also used for dedicated print jobs in a computer network and for routing jobs to printers and plotters.
• Network access server: A rack-mount server can be used to manage access to data, programs or networks in the form of a network access server.
• Storage: Depending on the scope and equipment of rack servers, you can use ad­di­tion­al storage resources or con­sol­id­ate them with other com­pon­ents.

Other areas of ap­plic­a­tion include:

• Proxy server
• DNS server
• Streaming server
• Gaming server
• Mail server

Rack servers differ in their features and di­men­sions. Racks with the 19-inch stand­ard­isa­tion are used most fre­quently and are usually divided into 2-post racks and 4-post racks. Depending on the width, several server com­pon­ents can be mounted ho­ri­zont­ally or ver­tic­ally in the racks. Vertical height units (rack units) are another important dis­tinc­tion. These have the stand­ard­ised dimension of 1.75 inches (4.445 cen­ti­metres). By default, racks comprise 42 height units.

Since not all slots are equipped from the start, blind panels are also used, i.e. place­hold­ers that provide a better ap­pear­ance, for example. Basically, the choice of rack server and com­pon­ents depends on the in­di­vidu­al per­form­ance and space re­quire­ments, because this also de­term­ines the required height and in­stall­a­tion depth

Trends toward more powerful, efficient data centres and growing demand for computing power and per­form­ance are driving server hardware to become ever more compact and smaller. Although rack servers are among the most widely used server tech­no­lo­gies, rack al­tern­at­ives such as blade servers and cartridge servers are catching on. These offer more mobility, save space and power, and enable easy scaling and con­fig­ur­a­tion through modular server com­pon­ents.

Micro servers are in­creas­ingly being con­sidered as al­tern­at­ives. However, classic tower servers are another option for small companies without server rooms and demands for high computing power.

The first com­mer­cial blade server was launched in 2001 and optimized the server tech­no­logy of rack servers. Blade servers cir­cum­ven­ted a major weakness of rack servers – the space problem caused by many server com­pon­ents in fixed rack-mount frames. Blade servers use a modular chassis with key com­pon­ents in the form of multiple, ultra-thin server blades. This means companies benefit from smaller server hardware with the same or greater computing power. Expansion and main­ten­ance are less com­plic­ated, as new com­pon­ents are simply slid in or out of the chassis.

However, blade servers are sig­ni­fic­antly more expensive to purchase, configure and set up. The re­place­ment of com­pon­ents can also be more complex due to a vendor lock-in. Compared to rack servers, blade servers, which are mostly used for dedicated ap­plic­a­tions, allow for fewer ex­ten­sions. That’s because the con­nec­tions for ad­di­tion­al hard disks or PCI slots are usually reduced. Therefore, rack servers are still the first choice for many companies due to their price-per­form­ance ratio.