Windows server network load balancing technology

Basic concepts of clusters

The Windows server cluster technology is a three-tier ApplicationProgramEach layer has a corresponding solution. For example, the network load balance (NLB) introduced in this article is a front-end cluster technology, and a COM + Component Load Balance (Component Load Balance, CLB), and the backend Service server cluster (Microsoft Cluster Server, MSC ).

Before introducing NLB, we should first understand some basic concepts related to clusters: What is a cluster and the benefits of using clusters.

In short, a cluster refers to a group of servers that provide the same service or application through mutual collaborative jobs to improve the availability of services or applications (availability, note 1) reliability (note 2) and scalability (note 3 ). When the server in the cluster fails, the request will be transferred to other servers to provide services that will not be interrupted 24 hours a day for the whole year.

Windows Server supports three cluster technologies: NLB, CLB, and MSC. NLB and MSC are built in Windows Server, and CLB needs to purchase application center.

The cluster technologies are described as follows:

1. NLB provides network traffic load balancing for TCP/IP-based services and applications to improve system availability and scalability. Common applications include Terminal Service, Web Service, and Web application.

NLB provides external services through a virtual IP address. When a request is received, NLB randomly determines who will process the request.

2. CLB provides dynamic load balancing for intermediary layer applications using COM + components to improve system availability and scalability. CLB determines who handles service requests based on the current workload.

3. msc provides Fault Tolerance transfer between backend services and applications to improve system availability. Common applications include SQL Server and Exchange Server.

The client determines who processes the service request. All servers share a share storage to store the session status. When the master server fails, it will continue to be taken over by the secondary server. The secondary server extracts the session status from share storage and continues the unfinished work to achieve Fault Tolerance transfer.

How NLB works

After having a basic understanding of the Windows server cluster, we will further understand how NLB distributes client service requests to servers in the cluster to distribute the traffic of service requests, so that the load is not concentrated on a server.NLB is used to filter pattern similarityAlgorithmTo allocate service requests. It is further divided into three types: no similarity, single similarity, and class C, which are described as follows:

1. No Similarity

Client Service requests are evenly distributed to each server in the cluster. Assume that the NLB cluster has two servers. When a client request is received, NLB submits 1st requests to 1st servers, 2nd requests to 2nd servers, and 3rd requests to 1st servers,... And so on. Because all client connections are evenly distributed to each server, the best load balancing can be achieved. To share the session state, you must store the session state in State or database server. This method applies to most applications.

2. Single Similarity

Client Service requests are allocated to a server in the cluster. When receiving a request from the client, NLB determines which server to handle Based on the client's IP address, that is, a server will only process requests from some IP addresses. Because the service requests of an IP address are only handled by one server, there is no session State sharing problem, but load imbalance may occur. This method is applicable to connections that require SSL (Secure Socket Layer) and multiple online communication protocols (such as FTP and PPTP ). SSL (Secure Socket Layer) is a set of encryption technologies that provide identity authentication, confidentiality, and data integrity. SSL is most commonly used to establish secure communication channels between Web browsers and web servers. It can also be used between client applications and Web services. PPTP (Point-to-Point Tunneling Protocol) is a network technology that supports Virtual Private Networks with multiple protocols.

3. class C is similar to a single similarity. The difference is that class C decides which server to handle Based on the blocked Class C of the IP, that is, a server only processes requests from some CIDR blocks C. This method ensures that clients using multiple proxies can direct to the same server.

Conclusion

Before creating NLBNote some restrictionsFor example, the private IP address of the server must be fixed and be in the same CIDR block. If there is only one Nic, the servers in the cluster cannot communicate with each other. Therefore, we recommend that you have two NICs, one for NLB and the other for heart beat between servers, checks whether the servers in the cluster are operating normally. If you use two NICs, you can only set the default gateway on one of the NICS. Otherwise, the packets cannot be forwarded to the correct IP address.

The basic concepts of NLB are briefly introduced to achieve some online settings of network load balancing administrators or regions using Windows server. For detailed steps, see "instructions and support ".

Note 1: Availability refers to the service levels provided by applications, services, or systems. High-availability systems rarely stop (either scheduled or unexpected ). Availability is usually expressed as the percentage of time when the service or system is available. For example, the Service Availability of 8.75 hours of downtime per year is 99.9%.

NOTE 2: Reliability refers to the possibility that a computer system or device can continue to operate within a specified period or under specified conditions.

NOTE 3: scalability is a way to measure the extent to which computers, services, or applications can expand to meet the increasing performance requirements. For a server cluster, when the overall load of the cluster exceeds its function, it can continuously add one or more systems to the existing cluster.

Reproduced from: http://www.seobbs.net/simple/index.php? T25276.html