Network boot process

This topic describes the process for Intel/AMD x86 computers, Intel/AMD x86-64 bit computers based on the UEFI standard, SUN SPARC and IBM® PowerPC® machines based on the OpenBoot standard.

A network boot application is different from any other program for a number of reasons:

• It is invoked by the BIOS during the initial bootstrap of the computer, before any operating system or disk boot manager.
• It depends on the presence of a special chip on the network card, the PXE boot ROM.
• It is capable of communicating with a server over the network thanks to a programming interface provided by the bootrom. PXE is the informal standard for this kind of programming interface.
• It is downloaded from a special OS deployment server, and does not depend on any local storage device. It can work on computers without hard disk, CD or diskette devices. If a local storage device is present on the computer, it is accessible to the network boot application. However, a network boot application does not fail if the storage device is corrupted or broken.
• It gets its parameters (IP address and other boot parameters) from a central DHCP server.

A typical network-boot sequence goes through the following phases:

1. Starting up: A user or a wake-up event starts the remote-boot computer up.
2. Network boot: The BIOS configuration (boot order), a hot key (for example, F12), or the wake-up event instructs the computer to boot on the network.
3. IP address discovery: The remote-boot target broadcasts a DHCP request for an IP address. Any DHCP server which knows the target (that is, recognizes its hardware address) or which has a pool of freely distributable dynamic addresses sends an IP address. The target takes the first answer and confirms it to the server. In addition to the IP address, the server gives some other network parameters to the target and information about the boot procedure to follow.
4. OS deployment server discovery: In the case of PXE remote-boot, the target computer then proceeds to the discovery of the OS deployment server. The OS deployment server is responsible for delivering a network boot program to the target. It is not necessarily the same computer as the DHCP server. The target responds to the first OS deployment server which replies and downloads a small program using a simple multicast protocol (MTFTP).
5. Server connection: If the network boot program is the deployment engine, it establishes a secure (encrypted) connection to the OS deployment server and receives instructions from the server to determine the name of the program to run.
6. Pre-OS configuration: The deployment engine then downloads from the file server everything required by the OS configuration specified by the system administrator. These file transfers are done using a secure, robust and efficient unicast or multicast protocol. Many actions can be performed at this stage: installing an operating system from scratch, repairing a corrupted operating system, or performing an inventory.
7. OS booting: When instructed by the OS configuration, the deployment engine removes itself from memory and allows the computer start the operating system, as if the target is booting normally from the hard disk. This ensures full compatibility with the operating system and avoids all problems of the traditional diskless remote boot.