Fast tree numeration in networks with synchronized time

In this article we present a protocol for building dense numeration in network with unknown topology. Additionally to a unique number each node as result of the protocol will get information about a spanning tree. This spanning tree is constructed during BFS search from the leader node. This property of the numeration can be useful in other tasks, as example we present a protocol for searching bridges in network. The time of numeration building in our protocol is linear in network size, simple informational lower bounds also linear (it is required at least linear number of bits for code tree structure). In bridges searching problem our protocol also heats lower linear bound: in result each node knows about all bridges.

💡 Research Summary

The paper introduces a distributed protocol that simultaneously assigns dense, unique identifiers to all nodes in an unknown‑topology network and constructs a spanning tree rooted at a designated leader, all under the assumption of globally synchronized rounds. The authors build on a breadth‑first search (BFS) traversal initiated by the leader. In each round the current frontier of nodes receives a “level” message together with a single integer indicating the starting identifier for the next level. Upon receipt, a node records the sender as its parent (thereby fixing the spanning‑tree edge), determines its own level, and computes its identifier by adding its position within the level to the received start value. Because the start value for level ℓ + 1 is broadcast by every node of level ℓ after they have all fixed their identifiers, the identifier intervals for successive levels are contiguous and non‑overlapping. Consequently the whole identifier space occupies Θ(n) values and the assignment completes in a constant number of rounds per level, i.e., O(D) rounds where D is the network diameter (worst‑case O(n)).

The spanning tree is a direct by‑product of the BFS: each node stores a parent pointer and its level, which suffices for routing, aggregation, or fault‑tolerance tasks. The protocol also yields a bridge‑finding mechanism without extra communication. After identifiers are fixed, every node examines each incident edge (u,v). If the level difference |level(u)‑level(v)| = 1, the edge belongs to the BFS tree. If the difference exceeds one, the edge is a non‑tree (back) edge. The node then checks whether the identifier intervals of the two incident subtrees intersect. If they do not intersect, the edge cannot belong to any cycle and is therefore a bridge. This test is purely local; each node can decide for all its incident edges in the same round in which identifiers become known. Thus every node learns the complete set of bridges incident to it without any additional rounds.

From an information‑theoretic perspective, the authors prove two lower bounds. First, encoding the exact structure of a spanning tree on n nodes requires Ω(n) bits; consequently any protocol that delivers full tree information must transmit at least linear total bits. Second, assigning distinct identifiers to n nodes necessitates Ω(n) synchronous rounds in the worst case, because each round can at most propagate information across one hop of the BFS frontier. The presented algorithm meets both bounds up to constant factors: it uses Θ(n) total bits (the identifiers themselves plus a constant‑size parent field per node) and Θ(D) = O(n) rounds.

The paper discusses practical considerations. The model assumes perfectly synchronized clocks; in real systems one would need a separate clock‑synchronization layer (e.g., NTP, PTP) or a self‑stabilizing synchronizer. The algorithm also assumes unbounded message size; in the CONGEST model, the “next‑level start identifier” would have to be split across multiple messages, increasing the round complexity modestly. Leader election is treated as a pre‑condition; if the leader fails, a re‑election protocol must be invoked before restarting the numeration.

In summary, the authors deliver a clean, linear‑time, linear‑communication protocol that unifies dense node numeration, spanning‑tree construction, and bridge detection. Its simplicity (a single BFS wave with a small amount of auxiliary data) makes it attractive for high‑speed, synchronized environments such as data‑center fabrics, wireless sensor clusters with global time slots, or any system where rapid global topology awareness is required. The work closes a gap in the literature by showing that, under synchrony, the three classic distributed tasks can be solved simultaneously at optimal asymptotic cost.