Interpretation of five mistaken ideas about Blade server

People are always willing to look at the new things with a critical eye, in the treatment of the blade server is so, some people for some repeatedly introduced technology ignored, still imposed on the blade of some trumped-up charges, these errors are affecting the promotion and application of Blade server.

misunderstanding one: the chassis, the plate will become "short board"

Now the blame for cooling and powering the blades has been reduced, and replacing the chassis and the plate has become a new target. At a CIO seminar at the beginning of this year, some users pointed out that the chassis, midplane (or backplane) will become the blade server "short board", because all the server Blade (Blade) and switch all connected/inserted in the plate, the plate once damaged, the consequences of unimaginable, and the chassis also has a similar problem. This concern, in fact, stems largely from subjective inference.

First of all, the chassis is a shell, its own damage probability is very small, there is nothing to worry about. The real concern is the blade server's medium board. IBM and HP are different in the design of redundancy in the middle board. As shown, this is the IBM BladeCenter Blade Server in the middle board, from the figure can be seen: in the upper and lower sides of the vent with 14 signals and power connectors, the IBM BladeCenter Blade Server full match is the 7u14 slice (7U height, 14 blades).

The HP BladeSystem Blade Server's middle panel does not use this up and down redundancy design, according to HP engineers, HP's plate and the server blade connectors, such as the connection does not have any electrical indicators, the connection between them is a pure plug in the connector between the electrical connection is through the controller to achieve, They are doubly redundant and can be replaced online. HP engineers pointed out that the bladesystem of course may also be damaged, but as a pure connectors, its probability is very low, this is after the laboratory fully certified conclusion, its life can reach hundreds of thousands of times. He pointed out that if the upper and lower redundancy design, once the side of the damage, or not to replace, so lose redundant protection, reliability has been reduced. If the replacement, only downtime, for users this is a dilemma choice.

　　 misunderstanding two: The standard flaw causes to be bound by the manufacturer

The issue of standardization has long been a hot topic. Several major blade server manufacturers, their specifications and standards are different, such as IBM provided BladeCenter S, E, H, T, HT and other 5 chassis, including the mainstream product is BladeCenter E, its configuration 7u14 tablets; H is a 9u14 slice, while H, HT is a chassis for telecommunications and high-performance design. HP Baldesystem C-class Series Blade server is divided into C3000 and C7000, which C3000 for small and medium-sized enterprises design, specifications 7U8 film; C7000 is a 10u16 piece. HP C-Series blades have a full height and a half height difference, there is a distinction between AMD and Intel blades for dual server products, with dual-core and four cores. The full high server occupies up to 2 slots, more Itanium or multiple servers, so far, can support 4-way dual-core or 4-way 4-core Xeon processor blades. In addition, HP also offers blade storage and tape drives.

Different specifications and systems, so that different manufacturers of server blades can not be mixed plug. From the perspective of the overall market, the lack of standards, not conducive to the formation of large-scale production, will affect the cost of blade server production. However, for users, this does not directly affect the blade server application, regardless of which blade server can be used to achieve rapid deployment, green energy saving and easy management purposes.

The user investment in the future can not be protected, such as users only buy a small number of blade servers, when the user's business reached a certain scale, perhaps 2-3 years later, users will not find the chassis compatible with the server blades, resulting in waste. In fact, this is also a trumped-up "guilt". For the size of the user, you can choose BladeCenter s or Baldesystem C3000, or simply select the domestic blade server, these blades in the design, fully considered the size and dosage of small and medium-sized enterprises. Taking Baldesystem C3000 as an example, the design is 7u8 film. 8 blades are not a large amount of for small and medium-sized enterprises, from the mail, file sharing, WEB, human resources, financial applications such as the need for multiple servers, in addition, some key applications, such as database applications need to use a full server, as well as storage, tape drives and other needs, so do not need to worry about the size of the problem

　　 Myth Three: Virtual connection is equivalent to virtualization

The virtual CONNECT,IBM of the blade server, known as virtual, enables similar functionality through the BladeCenter Open Fabric Manager software. The so-called virtual connection refers to when a server blade. For example, if a fails, replace it with a redundant d blade in the idle state. With the support of virtual connection, D can directly replace a, any related network, storage parameter configuration without reconfiguration, server blades can do "Plug and Play", Server blade and network, storage management becomes simple and transparent , this is the virtual connection.

The industry often confuse the virtual connection of the blade server with virtualization. It should be noted that the blade server does not unconditionally support virtual connections under any conditions. Only by using HP's network module (the blade server switch) can the blade server support the virtual connection. If a user chooses a Cisco, Brocade, or Foundry network module, the blade server does not support virtual connections unless the modules also have additional functionality.

　　 misunderstanding four: Blade change user Network structure

Because each server blade in the chassis is connected through a network module (switch), these network modules are also selectively available. However, compared to the rack server, between the server blade and LAN switch, after all, the addition of a network module (switch) layer, this will increase the difficulty of system management? In addition, the Blade Server network module will become a new bottleneck.

Technical engineers pointed out that in most cases, Blade server network module will not become a bottleneck, because it has a lot of choices, can support Gigabit Ethernet, can also support 10G, FC, InfiniBand, etc., its network module also has 2/3-layer switch function. In addition, for special applications, can also implement the blade server transmission. For example, the HP C Series provides an Ethernet pass-through module (pass through) that provides a straight-through LAN connection to the blade server. A one-to-one network adapter for wiring racks that provides non-blocking connectivity.

Blade Server This flexible networking capabilities, to provide users with more choices. In addition, the extension of the problem, such as the use of multiple network card connection, can also be extended to meet the requirements of the application.

　　 mistake five: Blade server cooling problem Outstanding

In the design of the blade server, heat dissipation is a key concern, but do not mistakenly think: Blade server will produce more heat, need to consume more air conditioning refrigeration. On the contrary, compared with the same number of 1U rack servers, the amount of refrigeration required by the blade is not increased, but is reduced by 25%.

This conclusion is not difficult to understand. In the interview, the product manager pointed out that, from the component, the blade server is not specific, it uses the processor, hard disk and memory, and the rack server is not different, there is no so-called to alleviate the heat pressure, and the use of laptop hard disk problem. In fact, blades in a blade server can be exactly equivalent to a rack-type server.

The blade server differs from a rack server in that it shares power supplies, fans, and chassis. In the power supply, the number of fans, blade server has been reduced, so more energy-saving, the resulting heat also decreased. Although different manufacturers adopt different heat dissipation schemes at present, they can meet the need of the blade server to dissipate heat. In addition, compared with the rack-type server, because the blade server has a relatively airtight space, so its cooling efficiency is higher.