Talking about how small and medium-sized enterprises view server performance design

With the rapid development of corporate business, enterprise information construction is gradually on the agenda. Enterprise informatization is an important means to promote the rapid development of small and medium-sized enterprises. Through the construction of enterprise informationization, it has a significant effect on the optimization of enterprise management mode, the standardization of business processes, and the improvement of work efficiency, so as to comprehensively improve the market of small and medium-sized enterprises. Competitiveness.

The basis of informationization of SMEs is the construction of information platform, and the focus of information platform construction is the choice of servers. As an important part of the entire information platform, the server is responsible for providing various types of network services, such as resource sharing and process control. It can be said that the performance of the server determines the success of information construction to a certain extent. This is also the case, in the process of server configuration, there are a variety of incorrect phenomena, such as in order to save money, just buy a server, after the work, there is a clear situation of small horse-drawn carts, resulting in information systems Not only can it not help the development of the enterprise, but it often drops the chain at a critical moment. Similarly, it has also cost a lot of money to buy high-end servers, which can be put into practice. In order to design a server performance solution suitable for small and medium-sized enterprises, and finally purchase a server suitable for the information system of the enterprise, we need to understand two aspects:

1. The main parameter indicators of server performance < Br>

1.1 Operation speed

The operation speed mainly includes the number of CPUs, word length, and main frequency. The number of CPUs has two meanings: one refers to the actual number of CPUs, and the other refers to the number of cores of a physical CPU. With the development of CPU technology, 4-core CPUs have begun to spread.

1.2 Memory System

The memory system includes memory capacity, speed, and support for memory mirroring. Because the server has a variety of technologies, it is much stricter than a compatible machine. It emphasizes not only the speed of memory, but also its inherent error correction technology capabilities and stability.

1.3 Disk Subsystem

In the server, the performance of the disk subsystem directly affects the overall performance of the server, especially in the database server and file server. Therefore, when selecting a server product, the disk subsystem became our second focus. The disk subsystem includes parameters such as disk capacity, RAID level, and disk IO throughput.

1.4 Network Subsystem

The network subsystem connects multiple computer systems with independent functions distributed in different locations by communication lines and communication devices, with the support of network software. A system that implements data communication and resource sharing between each other.
2, enterprise information system provides services and main performance requirements

At present, SMEs generally require services such as file services, database services, mail services, Web services, multimedia services, terminal services , other special business services, etc. Their requirements for the server are all different.

2.1 File Service

File Service is a relatively common information management application in enterprises, and it is also one of the more mature network applications. Such as FTP service, file sharing service, etc. In this type of service, the server is equivalent to a warehouse, ensuring fast transfer of data between the user and the server disk subsystem. The data is transferred from disk to memory and then transmitted to the client through the network. In this process, the speed of the network is generally Very slow, the impact on system performance is also the biggest, followed by the IO speed of the disk system, and the CPU and memory requirements are generally not high.

2.2 Database Services

The database has achieved tremendous development in the network age, database services are more and more widely used, and it is also the application with the highest performance requirements in all aspects of the system, such as finance and inventory. And personnel management applications. A high-performance CPU and a fast disk subsystem are required to satisfy a large number of random I/O requests and data transfers. Therefore, the bottleneck of database services to server performance is: memory, disk subsystem, and CPU.

2.3 Mail Service

With the development and popularity of the Internet, mail services are becoming an indispensable part of people's daily lives. The mail server plays the role of email router and warehouse. The bottleneck of server performance is: network subsystem, memory, disk subsystem and CPU.

2.4 Web Services

The performance of the Web server is determined by the content of the website. If the Web site is static, the system bottlenecks are: network subsystem and memory. If the web server is mainly doing intensive calculations (such as dynamically generating web pages). System performance bottlenecks are: memory, CPU, disk subsystem, and network subsystem.

2.5 Multimedia Services

Multimedia services are responsible for media control and media streaming on the network. I/O throughput plays a key role in server performance. The performance bottleneck of the video server is: network subsystem, disk subsystem and memory; audio service has low requirements on server hardware configuration, and the current server subsystem generally does not become a bottleneck.

2.6 Terminal Services

Terminal Services works by connecting the client and server to the standard LAN architecture via the TCP/IP protocol, executing various applications and transmitting the results to the user, all The load is added to the server. The bottleneck of system performance is usually: memory, CPU, network subsystem.

With the above two aspects, we can make a simple design of the server, but it can not accurately determine the performance indicators we actually need. For example, the same WEB service is provided, but the performance requirements of providing static WEB service and dynamic WEB service are obviously different, and the dynamic WEB service is also provided. The requirements on the server are different depending on the amount of calculation and the amount of access.
For SMEs, there may be cases where one or more services coexist. For example, when the author conducted an information platform for an import and export company, he found that the company has multiple service requirements:

(1) Internal business process control services.

(2) The need for file exchange within the enterprise.

(3) The need for enterprises to share the Internet.

(4) The need for corporate mail services.

In this case, relying on experience to complete the server performance parameter design is not enough, and the dedicated server in the enterprise can not make the correct design. Therefore, we generally use stress testing to identify bottlenecks in various subsystems of the server and to determine the number of users supported by the server.

There are many types of stress testing software, mainly including "system performance testing tools for server devices" and "testing tools for applications". The system performance testing tool for the server can perform special tests according to various subsystems such as processor, memory, storage, network, etc., and compare the same part of different servers through the test results to meet the performance requirements of some parts. service. Such as:

Iometer: storage subsystem read and write performance test

Sisoft Sandra: benchmark evaluation under WINDOWS

Iozone: I/O performance test under Linux

Netperf: Network Performance Tests

In addition, the "Test Tools for Applications" can check whether the server meets the expected design specifications under load and stress conditions. The goal of load testing is to test how the system's components, such as throughput, response time, CPU load, memory usage, etc., determine the performance of the system as the load increases. Such as:

Loadrunner: load test tool for predicting system behavior and performance

Webload: WEB performance stress test

In the case of import and export company information construction, first of all A simple analysis of the enterprise service type, select a dual-core, 1G memory, 100 megabyte network server, and then use Loadrunner to record the script, simulating the company's busiest time (40 employees) using proprietary software.

It turns out that the dedicated software does not support the dual-core. The multiple services cannot automatically allocate the CPU. As a result, one core of the dual-core CPU is 100%, while the other core is idle. When the number of users exceeds 20, the memory and disk exchange files are frequent, which seriously affects the efficiency of the server. At the moment of providing file services, there is a significant delay due to the problem of disk access speed and the dedicated service provided by the server. According to the problems found, we have redesigned the server and adopted virtualization technology to separate the database service, important service and secondary service, and configure different CPUs, memory and network resources respectively, on the basis of ensuring important services. , to meet other services.

According to the test results, and considering the future business development needs, you can configure dual-core dual-processor, 2G memory, 100-megabit network card server. After more than a year of practice testing, the server system is running well and can meet the user's performance requirements.

Of course, to choose a good server, there are often other parameters to be aware of, such as the operating system used by the server, the planning and allocation of resources, etc., but the design of the main performance parameters of the server is SME. The most basic and important factor in server performance design.