Although servers are different from personal computers yet in many ways server hardware is not too different from desktop hardware. Both are made of the same basic components like power supplies, memory, and CPUs. However, for those who have previously worked only with desktop hardware, server hardware can seem like a totally different ballgame altogether. Each component in server hardware tends to be more advanced than the ones found in personal computers. Servers sometimes also employ redundant and / or hot swappable components. Sometimes desktops too might have hot swappable components, but such characteristics are much more common on servers.
Towers look very similar to desktop PCs. If you are not concerned about space you can house more drives and more hardware into a single tower. However if you are looking at future expansion and scalability, space can become a constraint with a tower server.
A tower server is intended for use as a server and is built in an upright standalone cabinet. The cabinet called a tower is similar in dimensions to the cabinet of a tower-style desktop computer. Tower servers are different from rack servers or blade servers, which are generally rack-mounted. Tower servers are good as they provide easier cooling as the overall component density is comparatively low. The disadvantages include the fact that a set of tower servers is bulkier and weigh more than a blade server or set of rack servers. Also, the cabling for a big set of tower servers can be difficult and time-consuming. Finally, many air-cooled tower servers all placed in the same location can be noisy since each tower needs a dedicated fan.
Racks are designed to accommodate multiple servers, and are specially constructed to fit into small spaces. They are usually stacked right on top of one other, which makes cooling the devices difficult but is a big plus point when it comes to scalability. Also the stacked model helps in consolidating network resources and minimizing the floor space requirements.
In order to understand the hardware structure a bit better we need to understand what it consists of and how it is installed. A rack server, also known as rack-mounted server, is essentially a computer used as a server and is designed to be installed in a framework called a rack. The rack constitutes of multiple mounting slots known as bays, which are designed to secure a hardware unit held in place with screws.
Unlike the tower server, the rack server configuration simplifies cabling among network components. In an equipment rack filled with servers, a special cooling system is necessary to prevent excessive heat build-up that would otherwise occur when many power-dissipating components are confined in a small space.
Although blade servers are similar to rack servers in design, they are thinner, and also cost more than the latter. A blade server is a server chassis which houses multiple thin, modular EC boards, known as server blades. Each blade is actually a server, often dedicated to a single application. The server blades are literally servers on a card that contain processors, memory, integrated network controllers, and other input/output (IO) ports.
Ultimately, the type of server you choose depends on your needs. If space is your main concern, you can go for racks. If space is not a concern but cooling is then you can choose towers. And if money is not a concern and you want state-of-the-art server hardware which is space saving and works efficiently go for blade servers.