Fun Windows 2003 Load Balancing Policy

What is load balancing? Simply put, load balancing is to convert an IP address to another IP address, generally used for unregistered internal addresses and legal, registered Convert between Internet IP addresses. Load balancing not only effectively solves the performance limitations of a single server, but also enables rapid transfer of faults, ensuring high availability of services and flexible scalability. As a result, load balancing is almost a compulsory course for every network administrator. This article focuses on the application of Windows 2003 in Network Load Balancing.

Windows load balancing is mainly achieved by building a network load balancing cluster, supporting load balancing of servers such as WEB, FTP, Proxy, VPN, Windows Media, and Telnet.

The principle of load balancing

NLB is actually a series of servers that provide the same service to listen for service requests at the same time, and allows multiple application instances to run at the same time. The core of NLB is the filter driver for WLBS.SYS located between the network adapter driver and the network layer. NLB distributes each IP packet to all cluster nodes and processes it by one node according to the source address, destination address, transport layer protocol, port, cluster configuration parameters and algorithm of the packet, while other nodes drop the packet. Unified decision.

The concept of load balancing

Before configuring load balancing, you need to understand a few key concepts.

Cluster IP Address and Subnet Mask: The virtual IP address of the cluster, which is the "external" address that is presented to the customer.

Dedicated IP Configuration and Subnet Mask: The local IP address of each node in the cluster, uniquely identifying each node of the cluster.

Full Internet Name: The DNS name that can access this cluster, such as cluster.it.com.cn.

Cluster Operation Mode: Describes in the selection of the scheme.

Port Rules: The new features in Windows 2003, which refine the control of particles, can block the traffic of a particular application on a node, which is not possible in Windows 2000.

Priority (single host identifier): The range is between 1-32 (32 is the maximum number of nodes in a cluster). This value determines how incoming network traffic that is not included in any port rules defined for the cluster is processed. The host with the highest priority (minimum priority) will handle all such traffic.

Requirements for Load Balancing

Requirements for Operating Systems NLB is available in all versions of Windows 2003. The cluster is compatible with previous Windows server operating systems (such as 2000, NT4.0).

Network Architecture Requirements

NLB can run on servers connected to FDDI (Fiber Distributed Data Interface), Ethernet, Gigabit Ethernet, but not on Token Ring (Token Ring) Ring) runs on the network.

NIC Requirements

All network adapters must be in the Windows 2003 Hardware Compatibility List. In general, the server's network card will meet the requirements.

Switch and Router Requirements

When planning to use VLANs to prevent data flooding, you must ensure that the switch supports VLAN settings; when using multicast, some routers do not support unicast IP addresses. Map to a multicast MAC address, which needs to be set manually.

Requirements for Communication Protocols

Network adapters that are bound to a cluster can only be installed with TCP/IP protocol. They must be statically assigned and do not support DHCP.

Application Requirements

Must first be TCP or UDP communication, and make sure the current application or service must support NLB.

Load Balancing Design

Because network load balancing does not distribute traffic based on CPU and memory utilization, and performance does not vary linearly with the number of nodes (because scale The increase, the resulting network overhead, and CPU overhead also increase, so proper design and planning load balancing is critical.

The implementation of the cluster requires communication within the cluster (such as heartbeat information and aggregated communication) as well as data transfer for management and content replication. This part of the communication occupies the available bandwidth of the network. To overcome the limitations of a single NIC, you can use dual NICs, one for load client communication, and one for internal communication, management, and content data.

The choice of cluster operating mode is an important step in the design. Unicast mode means that each node's network adapter is reassigned a virtual MAC (consisting of 02-bf and cluster IP address to ensure the uniqueness of this MAC). Since the MAC addresses of all network adapters bound to the cluster are the same, in the case of a single NIC, the nodes cannot communicate with each other, which is one of the reasons for recommending dual NIC configuration. In order to avoid data flooding of the switch, it should be used in conjunction with VLAN. Previous12Next page Total 2 pages