In computer networks, out-of-band management involves the use of a dedicated channel for managing network devices. This allows the network operator to establish trust boundaries in accessing the management function to apply it to network resources. It also can be used to ensure management connectivity (including the ability to determine the status of any network component) independent of the status of other in-band network components.
In computing, one form of out-of-band management is sometimes called lights-out management (or LOM) and involves the use of a dedicated management channel for device maintenance. It allows a system administrator to monitor and manage servers and other network-attached equipment by remote control regardless of whether the machine is powered on, or if an operating system is installed or functional.
By contrast, in-band management like VNC or SSH is based on in-band connectivity and software that must be installed on the remote system being managed and only works after the operating system has been booted. This solution may be cheaper, but in computing it does not allow access to firmware (BIOS or UEFI) settings, does not make it possible to reinstall the operating system remotely, and it cannot be used to fix problems that prevent the system from booting. In networking, it does not allow management of remote network components independently of the current status of other network components.
Both in-band and out-of-band (OOB) management are usually done through a network connection, but an out-of-band management card can use a physically separated network connector if preferred. A remote management card usually has at least partially independent power supply, and can power the main machine on and off through the network.
This article focuses mainly on out-of-band management of servers, but also many (if not most) network devices offer out-of-band management. Modular/blade systems with dedicated management modules often offer a dedicated OOB Ethernet port or Lights out management port.
Capabilities
A complete remote management system allows[1] remote reboot, shutdown, powering on; hardware sensor monitoring (fan speed, power voltages, chassis intrusion, etc.); broadcasting of video output to remote terminals and receiving of input from remote keyboard and mouse (KVM over IP). It also can access local media like a DVD drive, or disk images, from the remote machine. If necessary, this allows one to perform remote installation of the operating system. Remote management can be used to adjust BIOS settings that may not be accessible after the operating system has already booted. Settings of hardware RAID or RAM clocking can also be adjusted as the management card needs no hard drives or main memory to operate.
As management via a serial port has traditionally been important on servers, a complete remote management system also allows one to talk with the server through this port (SOL console).
As sending monitor output through the network is bandwidth intensive, cards like MegaRAC use built-in video compression (versions of VNC are often used in implementing this). Devices like Dell DRAC also have a slot for a memory card where an administrator may keep server-related information independently from the main hard drive.
The remote system can be accessed either through an SSH command line interface, specialized client software, or through various web browser-based solutions. Client software is usually optimized to manage multiple systems easily.
There are also various scaled-down versions, up to devices that only allow remote reboot by power cycling the server. This helps if the operating system hangs but only needs a reboot to recover.
Implementation
Remote management can be enabled on many computers (not necessarily only servers) by adding a remote management card (while some cards only support a limited list of motherboards). Newer server motherboards often have built-in remote management and need no separate management card.
Internally, Ethernet-based out-of-band management can either utilize a dedicated separate Ethernet connection, or some kind of traffic multiplexing can be performed on the system's regular Ethernet connection. That way, a common Ethernet connection becomes shared between the computer's operating system and the integrated baseboard management controller (BMC), usually by configuring the network interface controller (NIC) to perform Remote Management Control Protocol (RMCP) ports filtering, use a separate MAC address, or to utilize virtual LAN (VLAN) tagging. Thus, out-of-band nature of the management traffic is ensured in a shared-connection scenario as the system configures the NIC to extract the management traffic from the incoming traffic flow on the hardware level, and to route it to the BMC before reaching the host and its operating system.
Remote CLI access
Another, older, version of out-of-band management is a layout involving availability of a separate network which allows network administrators to get command-line interface (CLI) access over console ports of network equipment, even when those devices are not forwarding any payload traffic.
If a location has several network devices, a terminal server can provide access to different console ports for direct CLI access. In case there is only one or just a few network devices, some of them provide AUX ports making it possible to connect a dial-in modem for direct CLI access. The mentioned terminal server can often be accessed via a separate network that does not use managed switches and routers for a connection to the central site, and/or it has a modem connected via dial-in access through POTS or ISDN.