weasel out of

Out-of-band management

In computing, out-of-band management (sometimes called lights-out management or LOM) involves the use of a dedicated management channel for device maintenance. It allows a system administrator to monitor and manage servers and other network equipment by remote control regardless of whether the machine is powered on.

By contrast, in-band management is the use of regular data channels (usually through Ethernet) to manage devices. A significant limitation of in-band management is its vulnerability to problems from the very devices that are being managed. To remotely manage network servers and routers, IT administrators need network access when problems occur. However, the same problems that cause the network to go down also result in the loss of management access to those devices.

Out-of-band management addresses this limitation by employing a management channel that is physically isolated from the data channel.

History

In the early 1980s, the concept of out-of-band was adapted for its natural application across the emerging data transmission network structures being introduced with the onset of Ethernet and cost-effective wide area networks. Network architects recognized that this out-of-band alternative pathway was a key requirement in service availability, and they could readily apply many of the lessons learned within the telecoms industry for the previous 30 years. Some of the earliest implementations of a data network Out-of-Band structure included the attachment of a single modem to any given server- in essence creating a very small Out-of-Band Infrastructure (OOBI). Vendors such as IBM, DEC, HP, and Data General made very lucrative service businesses by providing such out-of-band tools as subscription-based services products. The ability to remotely interact with servers that were otherwise compromised was dramatic and gave rise to the growth of Out-of-Band as a tool for data networks.

In the mid-1980s, Encore Computer released the Annex terminal server, later purchased by Xylogics. The Annex was capable of serving one parallel printer and up to 16 serial printers, terminals, modems, or serial consoles to various equipment and later models supported more serial lines; making it the first OOB management server. It also supported a "reverse telnet" (aka. rtelnet) feature that through a daemon created a character device file on the Unix host where it ran. Opening this device created a connection to the pre-configured port on the Annex, thus supporting remote kernel debugging, remote modems, etc.

Beginning in the year 2000, the concept was formalized by an early Out-of-Band Infrastructure for data pioneer Cyclades Corporation. It was quite clear that this technology was quickly becoming a core IT requirement when dealing with service-levels across hundreds or thousands of geographically dispersed IT assets. OOBI as it has been coined by Cyclades, uses many of the same concepts and provides similar features to the telecom industry's Out-of-Band Infrastructures. Vendors of OOBI solutions began offering these cost-effective alternatives to local administration for data system and network management. Just as in the past, a data OOBI provides alternate paths into the production infrastructure for the purpose of allowing disconnected assets to be remotely reconnected and subsequently returned to normal operation, in most cases eliminating the need for costly local administration. Some OOBI implementations include inherent enterprise-class security while others are constrained to the attributes of limited or proprietary mechanisms. An OOBI can improve operational efficiencies, cut costs, improve productivity and, in many cases, improve service levels and asset availability. Conceptually, data OOB infrastructures virtually guarantee a data dial-tone.

Types of management systems

A complete LOM system consists of a hardware component called the LOM module and a program that facilitates the continuous monitoring of variables such as microprocessor temperature and utilization. The program also allows for such remote operations as rebooting, shutdown, troubleshooting, alarm setting, fan-speed control, and operating system reinstallation. The program often integrates into traditional infrastructure in-band management tools such as HP Openview, Computer Associates, BMC, and Tivoli.

The most common out-of-band management solution involves connecting each device's serial console port to a console server. This implementation allows the monitoring of hardware self-test information and console access that is not available using typical in-band management.

Another type of management solution, a remote access card (RAC), involves an expansion card for a computer which has its own processor, memory, battery, network connection, and access to the system bus.

Some LOM systems function with more than one server, especially if combined with a KVM. When combined with a terminal server, administrators may access all serial console ports in a network or server farm from a single station. If the terminal server is also configured with network, Internet, and dial-up access, administrators will be able to manage network problems from any remote location, even if the network connection has been lost.

Communication between the controller and the remote servers sometimes takes place through an independent dial-up connection. More commonly nowadays, the LOM modules are connected by serial links to a separate management host; or the LOM module accepts telnet connections over an Ethernet connection. Either way, the LOM can then be remotely accessed over the Internet (through SSH to the management host, and/or a VPN). The LOM module keeps a record of all the operations (known as the event log), allowing the administrator to instantly check any or all of several hundred systems.

Console redirection

Embedded firmware of most server motherboards support (BIOS) serial-console redirection. And boot parameters of modern operating systems can be changed through the boot loader console which supports redirection as well (in Linux this is LILO, GRUB, or SYSLINUX). MS-Windows features EMS. Furthermore Unix-like systems can be configured to log kernel messages to their (serial) console too. Management software such as the Conserver automatically captures this data, and can replay it if needed. When using serial console servers care should be taken not to send any unsolicited BREAK over the line (especially with Sun hardware, and also Linux if SysRq is enabled) as it can put the machine in "lom mode" otherwise.

Limitations

Servicing and managing computer servers in a remote data center can require the physical presence of a system administrator. For example, the loading or removal of media, or direct interaction with the server through a console and keyboard (which should only ever be needed if the CMOS NVRAM becomes corrupted). Such access requirements depend on a system administrator being co-located with the data center, an additional business expense.