A space is 0-connected if it is in one piece (path-connected).
But the combinatorial layer held. The input complex—a twisted 12-dimensional shape of uncertainty—was subdivided, colored, and mapped via a simplicial approximation to the output complex of four regions. The satellites didn't agree on the exact vector. They agreed on the simplex of possible vectors.
A represents the collective state of processes.
: If two processors can start with either 0 or 1, the input complex forms a connected graph (a 1-dimensional complex) joining the states (0,0), (0,1), and (1,1). It has no holes; it is a single connected path. distributed computing through combinatorial topology pdf
He grabbed a napkin. "Combinatorial topology gives us the exact number. For 12 nodes and 3 failures, the minimum number of clusters we must allow is 4. That’s not a guess. That’s a homotopy invariant ."
If two processes execute concurrently, they cannot know who went first, creating a region of uncertainty.
The primary power of this approach is proving . If a mathematical "map" cannot be drawn from the starting shape to the ending shape without breaking certain topological rules, then no algorithm can solve that problem. A space is 0-connected if it is in
Through the lens of topology, an asynchronous execution creates "holes" in the state space.
Trapping an execution inside the hole means the processes cannot agree on a valid set of outputs.
In a standard wait-free shared-memory model where processes communicate via atomic read/write registers, executions can be modeled using immediate snapshots . When a set of processes execute a step, they write their current state and immediately read the states of all active processes. The satellites didn't agree on the exact vector
) : Represents all possible execution paths and intermediate states reachable by an actual protocol. The Topological Impossibility of Consensus
This part deals with general, "colored" tasks, which are more complex. It covers the general task solvability theorem and explores the topological structure of protocol complexes. Key topics include the (a simulation technique that reduces the number of faults) and a deeper exploration of the asynchronous computability theorem , which provides a complete characterization of tasks solvable in asynchronous systems.
A discrete version of the Brouwer Fixed-Point Theorem used to prove that at least one "winning" state must exist in certain protocols.
The definitive reference for this field is the book by Maurice Herlihy, Dmitry Kozlov, and Sergio Rajsbaum (2013). Distributed Computing Through Combinatorial Topology