Linux VS NT Server 4.0 's DHCP competition

DHCP (Dynamic Host Configuration Protocol) is the short name of dynamical host Configuration Protocol, which is evolved from BOOTP protocol, which belongs to TCP/IP protocol family, and is a protocol that uses more automatic client TCP/IP configuration on Ethernet at present. As a widely used network operating system, NT Server 4.0 and Linux provide DHCP support, this article is based on NT Server 4.0 with the DHCP server and Linux under the most common DHCPD 2.0 as an example to analyze the use of DHCP and the difference between the two systems.

The working principle of the DHCP protocol

Each TCP/IP protocol based network host (host) requires a uniquely identified 32-bit IP address to communicate with the network. If the IP address of each host is set by hand, the workload will be unimaginable, can you let the host automatically configure the IP address to function when it starts? This is the original intention of the DHCP protocol design. A DHCP-configured host is forced to send a limited address broadcast when it is started (a 32-bit full 1 IP address, known as a 255.255.255.255, is called a limited broadcast address, and broadcasts to this address are called limited address broadcasts or local webcasts). When a DHCP server in the local network receives a broadcast, it finds the appropriate configuration on the server based on the physical address received (physical addresses) and sends an IP address and additional options (such as lease expiration time, etc.) from the assigned IP pool to the client, due to the use of broadcasts, DHCP with no relays is clearly not able to cross subnets, and it is possible that a client's information is responded to by several servers, so the client always answers the first DHCP response received, and the client sends a registration message to tell the server that the IP has been leased to prevent an IP address conflict. The entire registration process is actually a fairly complex set of procedures. The two sides have to do multiple exchanges of information in order to finally register successfully.

Second, NT: graphical DHCP management

NT has always been known for its good operating interface, DHCP server is no exception, through the graphical DHCP manager, you can visually obtain the current DHCP client IP usage, you can easily modify the entire service project and client configuration. The NT DHCP server supports a considerable number of DHCP options, but Windows clients (including NT itself) only support fewer options, such as IP address, subnet mask, default gateway, DNS server name, domain name, WINS server name, WINS node type, lease duration, But this has covered most of the configuration in the LAN.

The NT DHCP server provides the client with an available IP, subnet mask, default gateway by defining one or more IP address ranges, but only one IP address range within a subnet. You can then provide more additional settings for the client with the DHCP option, and NT's additional settings are consistent with the RFC1541, such as 006 is the DNS server name, 044 is the WINS server name, and so on. If a DHCP client leases an IP address, the additional options are passed to it, provided that the client does not configure the additional options because the native configuration always takes precedence.

If you need to give some host-specific IP instead of automatic selection, you can add a reserved location and enter the 48-bit physical address of the host's Ethernet card as a unique identifier, as well as a specific configuration for those specific hosts. Whenever a client with this physical address logs on, the server gives it a fixed IP address and option.

The advantage of the NT 4.0 DHCP server is that it is intuitive and easy to manage, and one interesting place is that Windows clients always prioritize NT DHCP services (and the previous answer to the first response seems somewhat contradictory, but the truth is, In a subnet of all Windows hosts, Linux DHCP servers are rarely used if there are NT DHCP servers. Its main flaw is that it cannot effectively control the leased IP of unknown users, that is, as long as there are unused addresses, any requests received will be accepted, and it will still be cumbersome to configure more clients separately.