Server fleet management is software that lets a small team run thousands of servers as if they were one machine. It keeps a list of every server, updates their firmware, sets up new servers from templates and warns the team when a server is failing. Server makers sell it with their servers, and the largest cloud companies write their own. It is also called server management software or systems management.
Why is it needed? A large company may own several thousand servers in more than one data centre. Logging in to each server's baseboard management controller (BMC), the small computer inside it that controls it remotely, would take weeks. Fleet management software connects to every BMC over the management network instead. The team gives one instruction, and the software carries it out on every server.
It does four main jobs. First, it keeps an inventory: the model, serial number, parts and firmware version of every server. Second, it applies updates. Server makers regularly release new versions of firmware that fix security flaws. Third, it sets up new servers from a template, so that each new server gets the same settings as the hundred before it. Fourth, it collects health alerts, such as a failing disk or a fan running too fast, and passes them to the team's infrastructure monitoring tools.
Take, for example, the IT team of a large insurer in Gurugram, which looks after 2,400 servers in two data centres. A server maker announces a security flaw in its firmware. The software searches its inventory and finds 312 servers still running the affected version. The team schedules the update overnight, in batches of 50 servers, so that the insurer's apps never lose more than a small share of their servers at once. For each batch, the apps' work is first moved to other servers. The software then updates the firmware through each BMC, restarts each server and checks the new version number. By 5 a.m., 309 servers are updated. Three did not restart, and the software lists them for an engineer to check in the morning.
How is this different from configuration management? Fleet management looks after the hardware and its firmware, below the operating system. Configuration management sets up the operating system and the software above it. Large teams use both. A firmware update should be trusted only if the maker has signed it. On many new servers, a hardware root of trust checks that signature before the server starts.