Comprehensive analysis of energy management in cloud computing data center

The primary load in the data center is the server. Server development today, adapt to a variety of different functions, different environments of the server constantly appear, classification standards are also diverse.

Server classification

The primary load in the data center is the server. Server development today, adapt to a variety of different functions, different environments of the server constantly appear, classification standards are also diverse.

1. Grouped by application hierarchy into entry-level servers, workgroup-level servers, departmental servers, and enterprise-class servers

(1) Entry-level server

Entry-level servers typically use only processors, configure the appropriate memory and High-capacity IDE drives as needed, and use IDE RAID (a disk array technique, primarily to ensure data reliability and recoverability) for data protection, if necessary. Entry-level servers are designed for users of network operating systems, such as Windows NT ' NetWare, to meet the needs of office-type small and medium-sized network users for file sharing, print services, data processing, Internet access, and simple database applications. can also be completed in a small scope such as e-mail, Proxy, DNS and other services.

(2) Workgroup-level servers

Workgroup-level servers typically support 1 to 2 processors, which support large-capacity ECC (a memory technology, more for server memory) memory, full-featured. Highly manageable and easy to maintain, features required for small servers, such as SCSI (a bus interface technology) bus I/O (input/output) system, SMP symmetric multiprocessor architecture, optional raid, hot-swappable hard drive, hot-swappable power supply, etc., with high availability characteristics. Suitable for small and medium-sized enterprises to provide web, mail and other services, can also be used in schools and other educational departments of the Digital Campus network, multi-media classroom construction.

(3) Departmental Server

Departmental servers typically support 2 to 4 processors with high reliability, availability, scalability, and manageability. First, a large number of integrated monitoring and management circuits, with a comprehensive server management capabilities, can monitor such as temperature, voltage, fan, chassis and other state parameters. In addition, the combination of server management software, can enable managers to keep abreast of the server's work status. At the same time, most of the departmental servers have excellent system scalability, when users in the rapid increase in the volume of business can upgrade the system in time, can protect the user's investment. At present, the departmental server is a necessary link for the decentralized data acquisition units in enterprise networks to maintain a smooth communication with the highest level. Suitable for medium-sized enterprises (such as finance, telecommunications and other industries) as data centers, Web sites and other applications.

(4) Enterprise-Class Server

Enterprise-class servers are high-end servers, generally support 4 to 8 processors or more, with a separate dual-PCI channel and memory expansion board design, with high memory bandwidth, large-capacity hot-swappable hard disk and hot plug power, with strong data processing capabilities. This kind of product has the high fault-tolerant ability, the outstanding extended performance and the system performance, the extremely long system continuous running time, can protect the user's investment to a great extent. A database server that can be a large enterprise-level network. At present, enterprise-class servers are mainly applicable to the processing of large amounts of data, high processing speed and high reliability requirements of large enterprises and important industries (such as finance, securities, communications, telecommunications, telecommunications and other industries), can be used to provide ERP (enterprise resource allocation), E-commerce, OA (office automation) and other services.

2. According to the server's processor architecture (that is, the server CPU used by the instruction System) Division of the server into CISC architecture Server, RISC Architecture server and VLIW architecture server three kinds

Reference CISC Schema Server

CISC's English is all called "Complex instruction Setcomputer", or "Complex instruction system computer". Early desktop software is designed according to CISC, and has been continued to the present, so, microprocessor (CPU) vendors have been taking the CISC development path, covering Intel, AMD and so on. In the CISC microprocessor, the instructions for the program are executed sequentially in sequential order, and the operations in each instruction are serially executed sequentially. The advantage of sequential execution is that the control is simple, but the utilization of various parts of the computer is low and the execution speed is slow.

Article RISC Architecture Server

The English language of RISC is all called "Reduced instruction Setcomputing", the Chinese is "the simplified instruction set", its instruction system is relatively simple, it only requires the hardware to execute very limited and most commonly used part of the instruction, most complex operation uses the mature compilation technology, Synthesized by simple instructions. The CPUs that use RISC directives in the server are mainly HP's Alpha and PA-RISC, IBM's Power PC, MIPS's MIPS and Sun's redeemed.

(3) VLIW Schema Server

VLIW is an abbreviation for English "Very long Instruction Word", which means "long instruction set architecture", and the VLIW architecture uses advanced Epic (clear parallel instruction) design, which is also called "IA-64 architecture". The biggest advantage of VLIW is that it simplifies the structure of the processor, removed many complex control circuits within the processor, which are often used by superscalar chips (CISC and RISC) to coordinate parallel work, VLIW are simple in structure and can make their chip manufacturing cost lower, cheaper and less energy-efficient. And the performance is much higher than the superscalar chip. At present, the microprocessor based on this instruction architecture mainly has two kinds of were of Intel IA-64 and AMD.

3. Divided into general-purpose servers and dedicated servers by server use

(1) General-purpose server

A general-purpose server is a server that is not specifically designed for a particular service to provide a variety of service functions, and most servers are currently universal. This type of server is not designed for a particular function, so the design should take into account a variety of application needs, the server's structure is relatively complex, and requires higher performance, of course, the price is more expensive.

(2) Special-purpose server

A dedicated (or "functional") server is a server specifically designed for one or several functions. In some ways it is different from a general-purpose server. For example, a CD-ROM mirror server is used primarily for storing optical image files, so it requires a large capacity, high speed hard disk, and disc mirroring software. FTP servers are used primarily for file transfers on the Web, including the intranet and the Internet, which require the server to have a clear advantage in hard disk stability, access speed, I/O (input/output) bandwidth. and e-mail server is mainly to require the server to configure high-speed broadband Internet access tools, hard disk capacity and so on. These functional server performance requirements are relatively low, because it only needs to meet some of the needs of the functional applications, so the structure is relatively simple, the use of single CPU structure can be, in terms of stability, scalability and so on, the price is also cheaper.

4. Divided by server chassis structure, the server can be divided into "desktop server", "Rack Server", "Cabinet server" and "Blade server" four categories

(1) Desktop server

A desktop server is also known as a "tower server." Some desktop servers are similar to the size of ordinary vertical computer chassis, and some use large capacity of the chassis, like a huge cabinet. Low-grade server due to weak function, the entire server's internal structure is relatively simple, so the chassis is not large, all adopt desktop chassis structure.

(2) Rack-type Server

For the data center, select the server first to consider the size of the server, power, heat and other physical parameters, because the data center usually use large dedicated room to deploy and manage a large number of server resources, the room is usually equipped with strict security measures, good cooling system, multiple backup of the power supply. How to deploy more servers in a limited space is directly related to the cost of services to the enterprise, usually using a machine size to meet the 19-inch industrial standard rack-type server. Rack-mounted servers also have a variety of specifications, such as 1U (1u=1.75 inch =4.445cm), 2U, 4U, 6U, 8U, etc., installed in a standard 19-inch cabinet.

(3) Cabinet-type server

In some high-end enterprise servers due to the internal structure of complex, more internal equipment, and some have many different units or several servers are placed in a cabinet, this server is the cabinet-type server. For securities, banks, posts and telecommunications and other important enterprises, you should adopt a complete system of fault self-healing capabilities, the key components should be used redundant measures, for the key business use of the server can also be used for dual-machine hot backup high availability system or high-performance computer, such system availability can be very good guarantee.

(4) Blade server

The blade is a HAHD (high availabilityhigh density, highly available high-density) Low-cost server platform designed specifically for the particular application industry and high-density computing environments where each "blade" is actually a system motherboard, Similar to a separate server. In this mode, each motherboard runs its own system, serving a specified set of different user groups, with no association to each other. However, you can use system software to assemble these boards into a single server cluster. In cluster mode, all the motherboard can be connected to provide a high-speed network environment, you can share resources for the same user group services. There are two main types of blades in the current market: one for the telecom industry, the interface standard and the size specification in line with PICMG (PCI indus-trial Computer manufacturerscroup) 1.x or 2.x, and will be rolled out in line with PICMG 3. X of the products, the same standard of different manufacturers of blades and cabinets in theory can be mutually compatible; another type for the general calculation design, the interface may adopt the above standards or manufacturer standards, but the size specification is the manufacturer, pay attention to the performance price ratio.

(Responsible editor: The good of the Legacy)