One server is not like another. Depending on requirements and conditions, they occasionally vary heavily. From small desktop systems to rack-filling enterprise solutions there are different construction forms. And yet, the basic set-up is always similar despite the many options. One mainboard with processor, random access memory and external connection, cooling and power supply implemented in an arbitrary casing add up to a functioning machine. Should the need for computing power or memory change, individual components can be exchanged or upgraded. However, this is not always the most efficient way.
Especially fast growing oder variable structures are difficult to design and produce an overload or derating of the existent systems. And that is where the bladeserver-technology comes to use. A flexible and scalable solution with a very high power density that is in comparison easy to adjust to the requirements.
A bladeserver consists of essentially the same components like any other dedicated server. However, these are organised in two assemblies, whereby a certain modularity is achieved. These two assemblies are the chassis and serverblades.
The actual computers - the server blades – are composed of the main components of every server. The mainboard with the processors and random access memory as well as a network connection or LAN switch that is located with a small optional storage unit in a small slide-in case. The chassis on the other hand houses the (redundant) power supply, interfaces to the exterior as well as a module for administring the blade. Not until the assembly of the blade and chassis the bladeserver becomes a functioning bladeserver. The connection of the server blades inside the case takes place via a simple plug connection on the back plane of the chassis.
A single bladeserver can consist of a variety of single serverblades. Depending on manufacturer and construction, 10 to 24 blades per chassies are no problem. The processing units use the build in resources together and are centrally administrated.
Whilst the chassis is always the same basic set-up and only minor variations are possible, the range of differing server blades is bigger. They can be roughly divided into processing, memory and RAM units, which are then again alterable in details.
Computing- or processor blades provide the actual performance of the server. They are composed of processor and RAM by default and can be differently built in detail as needed. The main difference is the number of built in processors and their organisation. Established models support one to two processors or up to four processors per blade. A special form are the so called twinblades which unite two hardware-nodes per blade. This way you get two physical servers with two processors per blade.
While there is intended only limited space for storage in the processor blades and is not necessarily needed for operating, storage blades specialise in accomodation and administration of data. They less support the processing power and are in respect only minimally equipped regarding this. Instead, they provide space for more hard disks and support data management via e.g. RAID controller. Although with this construction form the space for theoretically 60 or more hard disks per bladeserver exists, the preferred version is an outsourced storage system or network.
The special forms primarily contain expansion capabilities via PCI slots. From assembling various additional controllers to supporting several graphic chips per blade, one can choose the optimal equipping for their purpose.
The smallest possible configuration of a working bladeserver theoratically consists of the chassis with one assembled blade. The complete equipping of the chassis is not required to establish functionality. In fact one of the big advantages is the possibility to leave leeway via spare slots. The blades themselves are all hot-swap capable. This means they can be exchanged in operation on the fly. The native management tool takes over the respective control and distribution of the computing load. Depending on approach, temporary failure of single components can occur though the bladeserver itself stays accessible the entire time.
Aside of that the bladeserver works the same way a classic server does. For example, it can be used for data management, hosting of websites or as communication system. There are no restrictions dependent on construction.
The primary advantages of the bladeserver technology are a high power density paired with high flexibility and scalability.
By the compact structure and the splitting of the ressources inside the chassis one can achieve a better performance per height unit in the rack. For instance, ten blades can be assembled in a seven unit high chassis. If we endow each blade with a twin module with two processors per node and equip the entire chassis with it, we count 20 hardware nodes, that is 40 processors. Accordingly we have reached a higher power density as it would have been possible with 1 HU rack server. So if one has high claims but little space at disposel, a bladeseserver may be worth a thought.
Herefrom results the first disadvantage of this technology: 20 hardware nodes consume a lot more power than 7 servers and produce a lot more waste heat as well. It is not for nothing that such a 7 HU-chassis can be equipped with four 3KW power adapters. So if you choose a bladeserver you would need a datacentre that can provide the needed cooling and power supply.
Operating a bladeserver cuts both ways. On one hand one saves space and a multitude of cables. For the power supply alone it is 20 or even 40 appliance inlets/connector plugs less since only the four built-in power adaptors in the chassis need to be plugged in. The same goes for the network connection. The internal communication ensues via the back plane of the chassis. Eventually a mouse/keyboard/grafic controller is enough to operate all implemented blades. Contrary to that is the sophisticated administration of the system itself because a 20 nodes consisting server is more complex to service than a classical 1 HU rack model.
Due to their modular structure, the bladeservers biggest advantage is their flexibility and scalability. With a rack server one settles for a model from the beginning. To change something subsequently results in a system failure due to the comprehensive rennovations. Or in the acquisition of new hardware.
Special purposes and demands or exceptional needs require individual hardware and hosting solutions. The Keyweb AG realises these in cooperation with the Keymachine Server Manufaktur GmbH for you. Within an intense consulting we suggest hardware according to your demands and get to the bottom of how/in what way a blade server supports your purpose, making it more efficient or sustainable. Or rather how it should be set up for it. After that, your custom made server will be exclusively build for you by the Keymachine Server Manufaktur GmbH. A Keymachine is created using verified, energy-efficient as well as stronng in performance hardware components only.
Choose features such as CPU or processor cores, working storage, memory and type of hard drive but also RAID options, backups operating system and administration software.
You can find more information to our dedicated custom made servers here.
With hosting in one of the TÜV-verified data centres of Keyweb you also decide for a multiredundant 80Gbit connection, USV supported power supply from 100% eco-power, entry-, fire-, and alarm protection as well as an intelligent air conditioning, technical emergency support 24/7/365 and highest data privacy under german law.
Profit also from our 19 years of experience. Our team will gladly answer all your questions under 0361 658 53 55 or info@keyweb.de.
