Competitive Capacitated Online Recoloring

In this paper, we revisit the online recoloring problem introduced recently by Azar, Machluf, Patt-Shamir and Touitou [Azar et al., 2022] to investigate algorithmic challenges that arise while scheduling virtual machines or processes in distributed systems and cloud services. In online recoloring, there is a fixed set V of n vertices and an initial coloring c₀: V → [k] for some k ∈ ℤ^{> 0}. Under an online sequence σ of requests where each request is an edge (u_t,v_t), a proper vertex coloring c of the graph G_t induced by requests until time t needs to be maintained for all t; i.e., for any (u,v) ∈ G_t, c(u)≠ c(v). In the distributed systems application, a vertex corresponds to a VM, an edge corresponds to the requirement that the two endpoint VMs be on different clusters, and a coloring is an allocation of VMs to clusters. The objective is to minimize the total weight of vertices recolored for the sequence σ. In [Azar et al., 2022], the authors give competitive algorithms for two polynomially tractable cases - 2-coloring for bipartite G_t and (Δ+1)-coloring for Δ-degree G_t - and lower bounds for the fully dynamic case where G_t can be arbitrary. We obtain the first competitive algorithms for capacitated online recoloring and fully dynamic recoloring, in which there is a bound on the number or weight of vertices in each color. Our first set of results is for 2-recoloring using algorithms that are (1+ε)-resource augmented where ε ∈ (0,1) is an arbitrarily small constant. Our main result is an O(log n)-competitive deterministic algorithm for weighted bipartite graphs, which is asymptotically optimal in light of an Ω(log n) lower bound that holds for an unbounded amount of augmentation. We also present an O(nlog n)-competitive deterministic algorithm for fully dynamic recoloring, which is optimal within an O(log n) factor in light of a Ω(n) lower bound that holds for an unbounded amount of augmentation. Our second set of results is for Δ-recoloring in an (1+ε)-overprovisioned setting where the maximum degree of G_t is bounded by (1-ε)Δ for all t, and each color assigned to at most (1+ε)n/(Δ) vertices, for an arbitrary ε > 0. Our main result is an O(1)-competitive randomized algorithm for Δ = O(√{n/log n}). We also present an O(Δ)-competitive deterministic algorithm for Δ ≤ ε n/2. Both results are asymptotically optimal.