2.5.7. Simulated Annealing¶
See also: The Annealer
Simulated Annealing is a modeling strategy from computer science that seeks optimal or stable outcomes through iterative analysis. A physical analogy for this is the process of strengthening steel by repeatedly heating, quenching and hammering. In both computer science and metallurgy, the process involves evaluating state, taking action, factoring in new data, and then repeating. Each annealing cycle improves the system, even though we may not know the final target state.
Annealing is well suited for problems where there is no mathematical solution, there’s an irregular feedback loop, or the datasets change over time. We have all three challenges in continuous data center operations environments. While it is clear that a deployment can be modeled as a directed graph (a mathematical solution) at a specific point in time, the reality is that there are too many unknowns to have a reliable graph. The problem is compounded because of unpredictable variance in hardware (NIC enumeration, drive sizes, BIOS revisions, network topology, etc) that prove more challenging if we factor in adapting to failures. Essentially, an operating infrastructure is a moving target that is hard to model predictively.
Digital Rebar implements the simulated annealing algorithm by decomposing the operations infrastructure into atomic units, known as node-roles (see Node Role APIs), that perform the smallest unit of work definable. Some or all of these node-roles are changed whenever the infrastructure changes. Digital Rebar anneals the environment by exercising the node-roles in a consistent way until system re-stabilizes.
One way to visualize Digital Rebar annealing is to imagine children who have to cross a field but do not have a teacher to coordinate them. One student takes a step forward and looks around. Another then sees the first and takes two steps. Each child advances based on what their classmates are doing. None of them want to get too far ahead or to be left behind. The line progresses irregularly but consistently based on the natural relationships within the group.
To understand the Digital Rebar Annealer, we have to break it down into three distinct components: deployment timeline, annealing, and node-role state. The deployment timeline represents externally (user, hardware, etc) initiated changes that propose a new target state. Once that new target is committed, Digital Rebar anneals by iterating through all the node-roles in a reasonable order. As the Annealer runs the node-roles, they update their own state. The aggregate state of all the node-roles determines the state of the deployment.
In Digital Rebar, a deployment is a combination of user and system defined state. Digital Rebar’s job is to stabilize the deployments and then maintain the desired configuration over time.