Load balancing more efficient - Windows Server 2003 implementation of traffic sharing
The server is the core of the enterprise information network. As the application is in-depth, the information flow in the enterprise network grows rapidly. When the amount of users and their applications are large, the host server at the same time may have to bear a large number of users' visits, but the processing power of a host is Limited, if the amount of visits exceeds the host's processing limit, the handling capacity of the host will become the bottleneck application of the enterprise network, which will restrict the continued development of network applications.
Solve the way
The traditional idea to solve this problem is that the replacement technology is more advanced, and the performance is more powerful, but this will enable the enterprise to face the huge pressure of capital investment, and for a growth company, the amount of network application information is very fast. The new host may encounter the same embarrassment situation in the future future.
Another solution is to use traffic sharing technology. That is to add multiple host servers in the network and allow these servers to save and process the same application content. Such host servers do not require the most advanced technology, the most powerful performance, so investment can be relatively small, but the server group consisting of them can work together to complete the network service function. When a user visits, these servers turn to the request of different users, and automatically disperse a large number of user requests into different host servers through the traffic sharing technology, thereby reducing the amount of tasks on a single host, achieving network traffic. Balance processing between the main unit.
The method of flow sharing technology is different depending on the hardware and software of the system, there is a general method, and some network systems provide specialized services. Microsoft's new released Windows Server 2003 provides many such technologies in its network, mainly distributed file system (DFS).
DFS
DFS is a distributed file system, which is mainly used to resolve problems that manage dispersed shared resource concentration. It is actually an important function, which is to use file replication service (FRS) in the domain environment to generate a copy of the shared directory.
The basic concept of DFS is DFS roots and DFS connections. DFS roots refers to the server or server group, which is where the client tries to access the file. These servers are typically distributed in each site in a domain. The DFS connection refers to a reference to the physical shared folder on the logical directory to the server that can handle file requests (anywhere within the enterprise). In Windows Server 2003, it greatly enhances the functionality of DFS. Windows Server 2003 has important functions for DFS in terms of reliability: A server can host multiple DFS roots. In Windows 2000, it is impossible to have multiple DFS roots on one server. Therefore, there is a need for a large number of servers running Windows 2000 to host multiple DFS roots. In Windows Server 2003, this restriction is canceled. In addition, the independent DFS server of the Windows 2000 cluster can only host a DFS root, while Windows Server 2003 allows you to host multiple DFS roots.
Windows Server 2003 also improves DFS behavior when setting up a copy operation across multiple sites. In Windows 2000, DFS will give priority to the target with the client in the same site, if there is no connection target in the site, it will request to select any connection target for the client in any other site. This algorithm is not most effective.
For example, if a company is located in Beijing, Tianjin and Shanghai, and its client computer tries to access the connection in Tianjin, the client computer does not consider communication costs when all the connection targets in Tianjin cannot access. The high and low, and the failover is transferred to a target in Beijing or Shanghai. In contrast, Windows Server 2003 uses site overhead information in Active Directory to select the site objective that meets the client request. For this situation, DFS can learn from the active directory to the communication cost from Tianjin to Shanghai than the communication fee from Tianjin to Beijing, so it will configure the corresponding client according to the station in the active directory. Redirect to the nearest Beijing goals. DFS
To achieve the flow sharing method, you must first maintain the same content in multiple host servers, that is, dynamically copy a host's specific content to multiple hosts. For example, in Windows Server 2003, we can do this with DFS technology. By using DFS in a domain environment, you can use a file copy service (FRS) to generate a copy of the shared directory, you can automatically copy the content under a folder in the specified host server to another or more servers. Thus, the purpose of maintaining the same content between the main unit is realized.
The first step in using DFS in Windows Server 2003 is to establish a DFS root directory.
From the Windows Server 2003 administrative tool to the Distributed File System Manager, create a new DFS root directory here. The system will require the root type when establishing a root directory. There are two types of DFS roots: one is domain DFS; the other is independent DFS, and if you want to implement a copy of the directory, you must select the domain DFS type. Independent DFS is only established in a non-domain environment, it cannot support file replication services.
To establish DFS in the domain environment, you need to specify the domain name and save the DFS root host server name, where the domain can also trust the domain. Then you want to define the name of the root as shown in Figure 1, this name is used to identify DFS roots in the domain. Then you want to specify a shared folder in the NTFS partition of the host server to place the DFS root. The second step is to establish a link. The link is a pointer to each shared directory within the network from the DFS root. On the Distributed File System Manager, use the mouse to use the DFS root that has been established and select New Link. In the wizard of the new link, you need to specify the target location to point to by the link, as shown in Figure 2, this location should be a shared directory already existing in a network. Then link a name of an image for this link.
Through the above-mentioned methods, each shared directory in the network can be organized in DFS, so that users can access all shared directories in it.
After the link of the domain DFS is created, you can create a copy. Specifies a shared directory on another server as a destination replication. Next, the specific object of replication is defined in the Copy Wizard. The shared directory in a host server can be specified in several other servers.
At this point, it is done in the work of the same content in multiple hosts. The DFS of Windows Server 2003 will use the file replication service to dynamically copy the specific content of a host server to multiple servers. By defining the properties of the DFS root, you can also post the DFS to the active directory, and when the domain users accesses the link content of this host server in the active directory, DFS will automatically deliver traffic in multiple servers. However, such applications only solve the access problem of the internal Windows users within the enterprise network, however, many non-enterprise network users may have access to the DFS root of the domain, such as the wide area network user to access the web page on the host server, is not directly accessed. The DFS root is therefore unable to access the link replica. Therefore, a more general method should also be selected to implement traffic sharing between different host servers.
DNS
For the above problems, it can be solved by the Domain Name Service (DNS). According to the work characteristics of DNS, our idea is to let DNS turn the same host name turnt to different IP addresses, that is, exchange the information to different network hosts. These hosts are our use of Windows Sever 2003 DFS. Copy technology is ready, the server saved the same content, thereby reaching the purpose of sharing network traffic between multiple hosts. Implement this feature can be done with the loop parsing function provided in the DNS server of Windows Server 2003.
First, "New Host" operations must be performed in the DNS manager, that is, specify the mapping correspondence of the host name and its IP address, which should be defined in the DNS server in the DNS server in the previous introduction of the previously introduced copy directory. . Attention should be noted, to specify different IPs to each host according to the network address of the host, but both hosts must play the same name. After this step definition, a host in the DNS server is configured with the same name but different addresses. Next, the server option is defined in the advanced configuration of the property setting. To reach the purpose of the DNS to swap the information to different hosts, you must select the Enable loop option here. This function is the key to this technology, which uses DNS to analyze the same host name turnt to different IP addresses.
After completing these settings, you can implement the multi-host server to share the network traffic in the Windows Server 2003 network. When many users accesses the network host, the DNS server will first respond to the first user first, and then respond to the second user's IP address to the second user, and so on. After the last IP address, the dynamic loop to the first host, push it at a subkey. The user does not need to know the specific network address of the host when access, as long as the name of the host can be accessed, all traffic is automatically completed by the DNS server, and the DNS server can also be dealt with by the main unit in the subnet according to the division of the subnet. User from the same subnet is accessed, so that network traffic is balanced between multiple host servers.
The method of this sharing flow is described above, using two services in Windows Server 2003, technology is completely implemented by software, and is relatively simple, and low cost, no additional hardware is required. Although the function is relatively small, the network technology of the operating system itself does not require additional other load balancing services, so this method is very versatile, any host server that provides directory copies and support for DNS loop parsing capabilities can be implemented.
In addition, in actual use, the number of hosts should be determined according to the traffic statistics of the network, and the number of hosts that can increase the host can increase the burden on file replication services and DNS services, but it is possible to reduce the overall performance of the system. For environments with high system requirements, complex environments, the program that can meet the application is selected based on the needs. In summary, the ultimate goal is to mitigate the load pressure of a single host server, but can not sacrifice the performance of the network, so the network is planned, such as correcting the subsidiary and determining the number of hosts in the subnet.
Ni
