On the complexity of distributed stable matching with small messages

We consider the distributed complexity of the stable matching problem (a.k.a. “stable marriage”). In this problem, the communication graph is undirected and bipartite, and each node ranks its neighbors. Given a matching of the nodes, a pair of unmatched nodes is called blocking if they prefer each other to their assigned match. A matching is called stable if it does not induce any blocking pair. In the distributed model, nodes exchange messages in each round over the communication links, until they find a stable matching. We show that if messages may contain at most B bits each, then any distributed algorithm that solves the stable matching problem requires ${\Omega(\sqrt{n/B\log n})}We consider the distributed complexity of the stable matching problem (a.k.a. “stable marriage”). In this problem, the communication graph is undirected and bipartite, and each node ranks its neighbors. Given a matching of the nodes, a pair of unmatched nodes is called blocking if they prefer each other to their assigned match. A matching is called stable if it does not induce any blocking pair. In the distributed model, nodes exchange messages in each round over the communication links, until they find a stable matching. We show that if messages may contain at most B bits each, then any distributed algorithm that solves the stable matching problem requires W(?{n/Blogn}){\Omega(\sqrt{n/B\log n})} communication rounds in the worst case, even for graphs of diameter O(log n), where n is the number of nodes in the graph. Furthermore, the lower bound holds even if we allow the output to contain O(?n){O(\sqrt n)} blocking pairs, and if a pair is considered blocking only if they like each other much more then their assigned match.     

A fault tolerant peer-to-peer spatial data structure

In recent years, many layered indexing techniques over distributed hash table (DHT)-based peer-to-peer (P2P) systems have been proposed to realize distributed range search. In this paper, we present a fault tolerant constant degree dynamic Distributed Spatial Data Structure called DSDS that supports orthogonal range search on a set of N d-dimensional points published on n nodes. We describe a total order binary relation algorithm to publish points among supernodes and determine supernode keys. A non-redundant rainbow skip graph is used to coordinate message passing among nodes. The worst case orthogonal range search cost in a d-dimensional DSDS with n nodes is \(O\left (\log n+m+\frac {K}{B}\right )\) messages, where m is the number of nodes intersecting the query, K is the number of points reported in range, and B is the number of points that can fit in one message. A complete backup copy of data points stored in other nodes provides redundancy for our DSDS. This redundancy permits answering a range search query in the case of failure of a single node. For single node failure, the DSDS routing system can be recovered to a fully functional state at a cost of O(log n) messages. Backup sets in DSDS nodes are used to first process a query in the most efficient dimension, and then used to process a query containing the data in a failed node in d-dimensional space. The DSDS search algorithm can process queries in d-dimensional space and still tolerate failure of one node. Search cost in the worst case with a failed node increases to \(O\left (d\log n+dm+\frac {K}{B}\right )\) messages for d dimensions.     

New solutions for disjoint paths in P systems

We propose four fast synchronous distributed message-based algorithms, to identify maximum cardinality sets of edge- and node-disjoint paths, between a source node and a target node in a digraph. Previously, Dinneen et?al. presented two algorithms, based on the classical distributed depth-first search (DFS), which run in O(mf) steps, where m is the number of edges and f is the number of disjoint paths. Combining Cidon??s distributed DFS and our new result, Theorem 3, we propose two improved DFS-based algorithms, which run in O(nf) steps, where n is the number of nodes. We also present improved versions of our two breadth-first search (BFS) based algorithms, with the same complexity upperbound, O(nf). Empirically, for a large set of randomly generated digraphs, our DFS-based edge-disjoint algorithm is 39?% faster than Dinneen et?al.??s edge-disjoint algorithm and our BFS-based edge-disjoint algorithm is 80?% faster. All these improved algorithms have been inspired and guided by a P system modelling exercise, but are suitable for any distributed implementation.     

Sleeping on the Job: Energy-Efficient and Robust Broadcast for Radio Networks

We address the problem of minimizing power consumption when broadcasting a message from one node to all the other nodes in a radio network. To enable power savings for such a problem, we introduce a compelling new data streaming problem which we call the Bad Santa problem. Our results on this problem apply for any situation where: (1) a node can listen to a set of n nodes, out of which at least half are non-faulty and know the correct message; and (2) each of these n nodes sends according to some predetermined schedule which assigns each of them its own unique time slot. In this situation, we show that in order to receive the correct message with probability 1, it is necessary and sufficient for the listening node to listen to a \(\Theta(\sqrt{n})\) expected number of time slots. Moreover, if we allow for repetitions of transmissions so that each sending node sends the message O(log?^? n) times (i.e. in O(log?^? n) rounds each consisting of the n time slots), then listening to O(log?^? n) expected number of time slots suffices. We show that this is near optimal.We describe an application of our result to the popular grid model for a radio network. Each node in the network is located on a point in a two dimensional grid, and whenever a node sends a message m, all awake nodes within L _∞ distance r receive m. In this model, up to \(t<\frac{r}{2}(2r+1)\) nodes within any 2r+1 by 2r+1 square in the grid can suffer Byzantine faults. Moreover, we assume that the nodes that suffer Byzantine faults are chosen and controlled by an adversary that knows everything except for the random bits of each non-faulty node. This type of adversary models worst-case behavior due to malicious attacks on the network; mobile nodes moving around in the network; or static nodes losing power or ceasing to function. Let n=r(2r+1). We show how to solve the broadcast problem in this model with each node sending and receiving an expected \(O(n\log^{2}{|m|}+\sqrt{n}|m|)\) bits where |m| is the number of bits in m, and, after broadcasting a fingerprint of m, each node is awake only an expected \(O(\sqrt{n})\) time slots. Moreover, for t≤(1?ε)(r/2)(2r+1), for any constant ε>0, we can achieve an even better energy savings. In particular, if we allow each node to send O(log?^? n) times, we achieve reliable broadcast with each node sending O(nlog?²|m|+(log?^? n)|m|) bits and receiving an expected O(nlog?²|m|+(log?^? n)|m|) bits and, after broadcasting a fingerprint of m, each node is awake for only an expected O(log?^? n) time slots. Our results compare favorably with previous protocols that required each node to send Θ(|m|) bits, receive Θ(n|m|) bits and be awake for Θ(n) time slots.     

An optimal bit complexity randomized distributed MIS algorithm

We present a randomized distributed maximal independent set (MIS) algorithm for arbitrary graphs of size n that halts in time O(log n) with probability 1 ? o(n ^?1), and only needs messages containing 1 bit. Thus, its bit complexity par channel is O(log n). We assume that the graph is anonymous: unique identities are not available to distinguish between the processes; we only assume that each vertex distinguishes between its neighbours by locally known channel names. Furthermore we do not assume that the size (or an upper bound on the size) of the graph is known. This algorithm is optimal (modulo a multiplicative constant) for the bit complexity and improves the best previous randomized distributed MIS algorithms (deduced from the randomized PRAM algorithm due to Luby (SIAM J. Comput. 15:1036?C1053, 1986)) for general graphs which is O(log² n) per channel (it halts in time O(log n) and the size of each message is log n). This result is based on a powerful and general technique for converting unrealistic exchanges of messages containing real numbers drawn at random on each vertex of a network into exchanges of bits. Then we consider a natural question: what is the impact of a vertex inclusion in the MIS on distant vertices? We prove that this impact vanishes rapidly as the distance grows for bounded-degree vertices and we provide a counter-example that shows this result does not hold in general. We prove also that these results remain valid for Luby??s algorithm presented by Lynch (Distributed algorithms. Morgan Kaufman 1996) and by Wattenhofer (http://dcg.ethz.ch/lectures/fs08/distcomp/lecture/chapter4.pdf, 2007). This question remains open for the variant given by Peleg (Distributed computing??a locality-sensitive approach 2000).     

An optimal maximal independent set algorithm for bounded-independence graphs

We present a novel distributed algorithm for the maximal independent set problem (This is an extended journal version of Schneider and Wattenhofer in Twenty-seventh annual ACM SIGACT-SIGOPS symposium on principles of distributed computing, 2008). On bounded-independence graphs our deterministic algorithm finishes in O(log* n) time, n being the number of nodes. In light of Linial’s Ω(log* n) lower bound our algorithm is asymptotically optimal. Furthermore, it solves the connected dominating set problem for unit disk graphs in O(log* n) time, exponentially faster than the state-of-the-art algorithm. With a new extension our algorithm also computes a δ + 1 coloring and a maximal matching in O(log* n) time, where δ is the maximum degree of the graph.     

An efficient distributed algorithm for maximum matching in general graphs

We present a distributed algorithm for maximum cardinality matching in general graphs. On a general graph withn vertices, our algorithm requiresO(n ^5/2) messages in the worst case. On trees, our algorithm computes a maximum matching usingO(n) messages after the election of a leader.     

Computing Associative Functions Distributively in Spite of Link Failures

Repeated computation of associative functions is central to many asynchronous distributed algorithms reported in the literature. We present efficient distributed algorithms for computing associative functions in spite of undetectable link failures in non-partitioned distributed systems. Two distributed; algorithms are presented for function computation assuming that the distributed system is preprocessed by a one-time preprocessing step that uses O(|E| |V|) messages (where |E| and |V| are the number of links and the number of nodes of the system, respectively). The first algorithm tolerates single link failures using O(|V| log |V|) messages and the second algorithm, which is an extension of the first algorithm, copes with the simultaneous failure of k links using O(k|E| log |V|) messages. Efficient computation of associative functions in the presence of multiple link failures has been an open problem, and our work solves this open problem.     

Distributed algorithms for ultrasparse spanners and linear size skeletons

We present efficient algorithms for computing very sparse low distortion spanners in distributed networks and prove some non-trivial lower bounds on the tradeoff between time, sparseness, and distortion. All of our algorithms assume a synchronized distributed network, where relatively short messages may be communicated in each time step. Our first result is a fast distributed algorithm for finding an ${O(2^{{\rm log}^{*} n} {\rm log} n)}We present efficient algorithms for computing very sparse low distortion spanners in distributed networks and prove some non-trivial lower bounds on the tradeoff between time, sparseness, and distortion. All of our algorithms assume a synchronized distributed network, where relatively short messages may be communicated in each time step. Our first result is a fast distributed algorithm for finding an O(2^{log* n} log n){O(2^{{\rm log}^{*} n} {\rm log} n)} -spanner with size O(n). Besides being nearly optimal in time and distortion, this algorithm appears to be the first that constructs an O(n)-size skeleton without requiring unbounded length messages or time proportional to the diameter of the network. Our second result is a new class of efficiently constructible (α, β)-spanners called Fibonacci spanners whose distortion improves with the distance being approximated. At their sparsest Fibonacci spanners can have nearly linear size, namely O(n(loglogn)^f){O(n(\log \log n)^{\phi})} , where f = (1 + ?5)/2{\phi = (1 + \sqrt{5})/2} is the golden ratio. As the distance increases the multiplicative distortion of a Fibonacci spanner passes through four discrete stages, moving from logarithmic to log-logarithmic, then into a period where it is constant, tending to 3, followed by another period tending to 1. On the lower bound side we prove that many recent sequential spanner constructions have no efficient counterparts in distributed networks, even if the desired distortion only needs to be achieved on the average or for a tiny fraction of the vertices. In particular, any distance preservers, purely additive spanners, or spanners with sublinear additive distortion must either be very dense, slow to construct, or have very weak guarantees on distortion.     

Space and Time Efficient Self-Stabilizing ℓ-Exclusion in Tree Networks

We propose an efficient self-stabilizing ℓ-exclusion algorithm in rooted tree networks running under an unfair distributed daemon. The ℓ-exclusion problem is a generalization of the mutual exclusion problem—ℓ (ℓ⩾1) processors, instead of 1, are permitted to use a shared resource. The algorithm is semi-uniform and its space requirement is (ℓ+3)Δ_r states (or ⌈log((ℓ+3)Δ_r)⌉ bits) for the root r, 4(Δ_p−1) states (or ⌈2 log(Δ_p−1)⌉ bits) for an internal processor p, and 3 states (or 2 bits) for a leaf processor, where Δ_p is the degree of processor p. This is the first ℓ-exclusion algorithm on trees with the property that the space requirement is independent of the size of the network for any processor, and is independent of ℓ for all processors except the root. The stabilization time of the algorithm is only O(ℓ+h) rounds, where h is the height of the tree.     

Local MST Computation with Short Advice

We use the recently introduced advising scheme framework for measuring the difficulty of locally distributively computing a Minimum Spanning Tree (MST). An (m,t)-advising scheme for a distributed problem ? is a way, for every possible input I of ?, to provide an “advice” (i.e., a bit string) about I to each node so that: (1) the maximum size of the advices is at most m bits, and (2) the problem ? can be solved distributively in at most t rounds using the advices as inputs. In case of MST, the output returned by each node of a weighted graph G is the edge leading to its parent in some rooted MST T of G. Clearly, there is a trivial (?log?n?,0)-advising scheme for MST (each node is given the local port number of the edge leading to the root of some MST T), and it is known that any (0,t)-advising scheme satisfies $t\geq\tilde{\Omega}(\sqrt{n})We use the recently introduced advising scheme framework for measuring the difficulty of locally distributively computing a Minimum Spanning Tree (MST). An (m,t)-advising scheme for a distributed problem ℘ is a way, for every possible input I of ℘, to provide an “advice” (i.e., a bit string) about I to each node so that: (1) the maximum size of the advices is at most m bits, and (2) the problem ℘ can be solved distributively in at most t rounds using the advices as inputs. In case of MST, the output returned by each node of a weighted graph G is the edge leading to its parent in some rooted MST T of G. Clearly, there is a trivial (⌈log n⌉,0)-advising scheme for MST (each node is given the local port number of the edge leading to the root of some MST T), and it is known that any (0,t)-advising scheme satisfies t 3 [(W)\tilde](?n)t\geq\tilde{\Omega}(\sqrt{n}). Our main result is the construction of an (O(1),O(log n))-advising scheme for MST. That is, by only giving a constant number of bits of advice to each node, one can decrease exponentially the distributed computation time of MST in arbitrary graph, compared to algorithms dealing with the problem in absence of any a priori information. We also consider the average size of the advices. On the one hand, we show that any (m,0)-advising scheme for MST gives advices of average size Ω(log n). On the other hand we design an (m,1)-advising scheme for MST with advices of constant average size, that is one round is enough to decrease the average size of the advices from log n to constant.     

Computing functions on asynchronous anonymous networks

In an “anonymous” network the processors have no identity numbers. We investigate the problem of computing a given functionf on an asynchronous anonymous network in the sense that each processor computesf(I) for any inputI = (I(v ₁),...,I(v _n )), whereI(v _i) is the input to processorv _i ,i = 1, 2,...,n. We address the following three questions: (1) What functions are computable on a given network? (2) Is there a “universal” algorithm which, given any networkG and any functionf computable onG as inputs, computesf onG? (3) How can one find lower bounds on the message complexity of computing various functions on anonymous networks? We give a necessary and sufficient condition for a function to be computable on an asynchronous anonymous network, and present a universal algorithm for computingf(I) on any networkG, which acceptsG andf computable onG, as well as {I(v _i )}, as inputs. The universal algorithm requiresO(mn) messages in the worst case, wheren andm are the numbers of processors and links in the network, respectively. We also propose a method for deriving a lower bound on the number of messages necessary to solve the above problem on asynchronous anonymous networks.     

Distributed algorithm for extrema-finding in circular configuration of processors

A distributed algorithm for finding the leader (highest numbered node) of a circular configuration of n processors is presented. Bounds on the number of message transmissions are derived. The average number of messages transmitted is of O(nlnn).     

Fast and proximity-aware multi-source overlay multicast under heterogeneous environment

Overlay multicast has been considered as one of the most important developments for the next generation Internet infrastructure. In this paper, we consider overlay multicast in the scenarios where any participant node is a potential data source. Existing multicast algorithms for single-source always require a long time to deliver messages or have high maintenance overhead when multiple data sources are allowed. There are other algorithms that are designed for multi-source scenarios. But they consume too much network resources and have a long convergence time because of proximity ignorance. To address the issues, we present FPCast, which leverages node heterogeneity and proximity information at the same time. Physically close nodes are grouped into clusters and each cluster selects a powerful, stable node as its rendezvous point. The rendezvous nodes form a DHT-based structure. Data messages are replicated and forwarded along implicit, source specific, and heterogeneity-aware multicast trees. We further reduce the delivery delay by introducing probabilistic forwarding scheme. We show the average delivery path length converges to O(logn) automatically (n is the number of nodes in the overlay). The simulation results demonstrate the superiority of our algorithm in terms of message delivery time and network resource consumption, in comparison with the previous randomized algorithms. The algorithm is also resilient to node failures.     

Efficient parallel algorithms forr-dominating set andp-center problems on trees

We develop efficient parallel algorithms for ther-dominating set and thep-center problems on trees. On a concurrent-read exclusive-write PRAM, our algorithm for ther-dominating set problem runs inO(logn log logn) time withn processors. The algorithm for thep-center problem runs inO(log² n log logn) time withn processors.     

An improved parallel algorithm for integer GCD

We present a simple parallel algorithm for computing the greatest common divisor (gcd) of twon-bit integers in the Common version of the CRCW model of computation. The run-time of the algorithm in terms of bit operations isO(n/logn), usingn ^1+? processors, where ? is any positive constant. This improves on the algorithm of Kannan, Miller, and Rudolph, the only sublinear algorithm known previously, both in run time and in number of processors; they requireO(n log logn/logn),n ² log² n, respectively, in the same CRCW model. We give an alternative implementation of our algorithm in the CREW model. Its run-time isO(n log logn/logn), usingn ^1+? processors. Both implementations can be modified to yield the extended gcd, within the same complexity bounds.     

Independent spanning trees on twisted cubes

Multiple independent spanning trees have applications to fault tolerance and data broadcasting in distributed networks. There are two versions of the n independent spanning trees conjecture. The vertex (edge) conjecture is that any n-connected (n-edge-connected) graph has n vertex-independent spanning trees (edge-independent spanning trees) rooted at an arbitrary vertex. Note that the vertex conjecture implies the edge conjecture. The vertex and edge conjectures have been confirmed only for n-connected graphs with n≤4, and they are still open for arbitrary n-connected graph when n≥5. In this paper, we confirm the vertex conjecture (and hence also the edge conjecture) for the n-dimensional twisted cube TQ_n by providing an O(NlogN) algorithm to construct n vertex-independent spanning trees rooted at any vertex, where N denotes the number of vertices in TQ_n. Moreover, all independent spanning trees rooted at an arbitrary vertex constructed by our construction method are isomorphic and the height of each tree is n+1 for any integer n≥2.     

Efficient and Scalable PRAM Algorithms for Discrete-Event Simulation of Bounded Degree Networks

We describe two new parallel algorithms, one conservative and another optimistic, for discrete-event simulation on an exclusive-read exclusive-write parallel random-access machine (EREW PRAM). The target physical systems are bounded degree networks which are represented by logic circuits. Employing p processors, our conservative algorithm can simulate up to O(p) independent messages of a system with n logical processes in O(log n) time. The number of processors, p, can be optimally varied in the range 1 ≤ p ≤ n. To identify independent messages, this algorithm also introduces a novel scheme based on a variable size time window. Our optimistic algorithm is designed to reduce the rollback frequency and the memory requirement to save past states and messages. The optimistic algorithm also simulates O(p) earliest messages on a p-processor computer in O(log n) time. To our knowledge, such a theoretical efficiency in parallel simulation algorithms, conservative or optimistic, has been achieved for the first time.     

Fault-tolerant routing over shortest node-disjoint paths in hypercubes

This paper presented a routing algorithm that finds n disjoint shortest paths from the source node s to target node d in the n-dimensional hypercube. Fault-tolerant routing over all shortest node-disjoint paths has been investigated to overcome the failure encountered during routing in hypercube networks. In this paper, we proposed an efficient approach to provide fault-tolerant routing which has been investigated on hypercube networks. The proposed approach is based on all shortest node-disjoint paths concept in order to find a fault-free shortest path among several paths provided. The proposed algorithm is a simple uniform distributed algorithm that can tolerate a large number of process failures, while delivering all n messages over optimal-length disjoint paths. However, no distributed algorithm uses acknowledgement messages (acks) for fault tolerance. So, for dealing the faults, acknowledgement messages (acks) are included in the proposed algorithm for routing messages over node-disjoint paths in a hypercube network.     

Computing Diameter in the Streaming and Sliding-Window Models

We investigate the diameter problem in the streaming and sliding-window models. We show that, for a stream of nn points or a sliding window of size nn, any exact algorithm for diameter requires W(n)\Omega(n) bits of space. We present a simple e\epsilon-approximation algorithm for computing the diameter in the streaming model. Our main result is an e\epsilon-approximation algorithm that maintains the diameter in two dimensions in the sliding-window model using O((1/e^3/2) log³n(logR+loglogn + log(1/e)))O(({1}/{\epsilon^{3/2}}) \log^{3}n(\log R+\log\log n + \log ({1}/{\epsilon}))) bits of space, where RR is the maximum, over all windows, of the ratio of the diameter to the minimum non-zero distance between any two points in the window.