Section: New Results

Theory of Distributed Systems

Simulation of Partial Replication in Distributed Transactional Memory

Participant : François Taïani.

Distributed Transactional Memory (DTM) is a concurrency mechanism aimed at simplifying distributed programming by allowing operations to execute atomically, mirroring the well-known transaction model of relational databases. DTM can play a fundamental role in the coordination of participants in mobile distributed applications. Most DTM solutions follow a full replication scheme, in spite of recent studies showing that partial replication approaches can present gains in scalability by reducing the amount of data stored at each node. This work [33] investigates the role of replica location in DTMs. The goal is to understand the effect of latency on the DTM's system performance in face of judicious replica distribution, taking into consideration the locations where data is more frequently accessed.

This work was performed in collaboration with with Diogo Lima and Hugo Miranda from the University of Lisbon (Portugal).

Distributed Universal Constructions: a Guided Tour

Participant : Michel Raynal.

The notion of a universal construction is central in computing science: the wheel has not to be reinvented for each new problem. In the context of n-process asynchronous distributed systems, a universal construction is an algorithm that is able to build any object defined by a sequential specification despite the occurrence of up to (n − 1) process crash failures. Michel Raynal presented a guided tour of such universal constructions in the bulletin of the EATCS [22]. Its spirit is not to be a catalog of the numerous constructions proposed so far, but a (as simple as possible) presentation of the basic concepts and mechanisms that constitute the basis these constructions rest on.

Atomic Read/Write Memory in Signature-Free Byzantine Asynchronous Message-Passing Systems

Participant : Michel Raynal.

This work introduced a signature-free distributed algorithm which builds an atomic read/write shared memory on top of a fully connected peer-to-peer n-process asynchronous message-passing system in which up to t<n/3 processes may commit Byzantine failures. From a conceptual point of view, this algorithm is designed to be as close as possible to the algorithm proposed by  [42], which builds an atomic register in an n-process asynchronous message-passing system where up to t<n/2 processes may crash. The proposed algorithm is particularly simple. It does not use cryptography to cope with Byzantine processes, and is optimal from a t-resilience point of view (t<n/3). A read operation requires O(n) messages, and a write operation requires O(n2) messages. This work was done in collaboration with Achour Mostéfaoui, Matoula Petrolia and Claude Jard from the University of Nantes and was published in Theory of Computing Systems [19].

From wait-free to arbitrary concurrent solo executions in colorless distributed computing

Participant : Michel Raynal.

In an asynchronous distributed system where any number of processes may crash, a process may have to run solo, computing its local output without receiving any information from other processes. In the basic shared memory system where the processes communicate through atomic read/write registers, at most one process may run solo.

In this work we introduced a new family of d-solo models, where d-processes may concurrently run solo, 1≤d≤n (the 1-solo model is the basic read/write model). We studied distributed colorless computations in the d-solo models, where process ids are not used, either in task specifications or during computation, and we characterized the colorless tasks that can be solved in each d-solo model. Colorless tasks include consensus, set agreement and many other previously studied tasks. This shows that colorless algorithms have limited computational power for solving tasks, only when d>1. When d=1, colorless algorithms can solve the same tasks as algorithms that may use ids. It is well-known that, while consensus is not wait-free solvable in a model where at most one process may run solo, ϵ-approximate agreement is solvable. In a d-solo model, the fundamental solvable task is (d,ϵ)-solo approximate agreement, a generalization of ϵ-approximate agreement. Indeed, (d,ϵ)-solo approximate agreement can be solved in the d-solo model, but not in the (d+1)-solo model.

This work was carried out in collaboration with Maurice Herlihy from Brown University, Sergio Rajsbaum from UNAM (Mexico), and Julien Stainer from EPFL, in the context of the LIDICo associate team. It was published in Theoretical Computer Science [18].

Early Decision and Stopping in Synchronous Consensus: A Predicate-Based Guided Tour

Participant : Michel Raynal.

Consensus is the most basic agreement problem encountered in fault-tolerant distributed computing: each process proposes a value and non-faulty processes must agree on the same value, which has to be one of the proposed values. While this problem is impossible to solve in asynchronous systems prone to process crash failures, it can be solved in synchronous (round-based) systems where all but one process might crash in any execution. It is well-known that (t + 1) rounds are necessary and sufficient in the worst case execution scenario for the processes to decide and stop executing, where t < n is a system parameter denoting the maximum number of allowed process crashes and n denotes the number of processes in the system. Early decision and stopping considers the case where f < t processes actually crash, f not being known by processes. It has been shown that the number of rounds that have to be executed in the worst case is then min(f + 2, t + 1). In this work we showed that this value is an upper bound attained only in worst execution scenarios. This work resulted from a collaboration with Armando Castaneda from UNAM, Yoram Moses from Technion, and Matthieu Roy from LAAS Toulouse, in the context of the LIDICo associate team. It was published at NETYS 2017 [29].

Long-Lived Tasks

Participant : Michel Raynal.

The predominant notion for specifying problems to study distributed computability are tasks. Notable examples of tasks are consensus, set agreement, renaming and commit-adopt. The theory of task solvability is well-developed using topology techniques and distributed simulations. However, concurrent computing problems are usually specified by objects. Tasks and objects differ in at least two ways. While a task is a one-shot problem, an object, such as a queue or a stack, typically can be invoked multiple times by each process. Also, a task, defined in terms of sets, specifies its responses when invoked by each set of processes concurrently, while an object, defined in terms of sequences, specifies the outputs the object may produce when it is accessed sequentially.

In this work we showed how the notion of tasks can be extended to model any object. A potential benefit of this result is the use of topology, and other distributed computability techniques to study long-lived objects. This work resulted from a collaboration with Armando Castaneda and Sergio Rasjbaum from UNAM in the context of the LIDICo associate team. It was published at NETYS 2017 [35].

Which Broadcast Abstraction Captures k-Set Agreement?

Participant : Michel Raynal.

It is well-known that consensus (one-set agreement) and total order broadcast are equivalent in asynchronous systems prone to process crash failures. Considering wait-free systems, we addressed and answered the following question: which is the communication abstraction that "captures" k-set agreement? To this end, we introduced a new broadcast communication abstraction, called k-BO-Broadcast, which restricts the disagreement on the local deliveries of the messages that have been broadcast (1-BO-Broadcast boils down to total order broadcast). Hence, in this context, k=1 is not a special number, but only the first integer in an increasing integer sequence. This establishes a new "correspondence" between distributed agreement problems and communication abstractions, which enriches our understanding of the relations linking fundamental issues of fault-tolerant distributed computing. This work was carried out in collaboration with Damien Imbs from the University of Marseille, Achour Mostéfaoui from the University of Nantes, and Matthieu Perrin from IMDEA (Spain). It was published at DISC 2017 [39].

Signature-free asynchronous Byzantine systems: from multivalued to binary consensus with t< n/3, O(n2) messages, and constant time.

Participant : Michel Raynal.

We introduced a new algorithm that reduces multivalued consensus to binary consensus in an asynchronous message-passing system made up of n processes where up to t may commit Byzantine failures. This algorithm has the following noteworthy properties: it assumes t<n/3t<n/3 (and is consequently optimal from a resilience point of view), uses O(n2) messages, has a constant time complexity, and uses neither signatures nor additional computational power (such as random numbers, failure detectors, additional scheduling assumption, or additional synchrony assumption). The design of this reduction algorithm relies on two new all-to-all communication abstractions. The first one allows the non-faulty processes to reduce the number of proposed values to c, where c is a small constant. The second communication abstraction allows each non-faulty process to compute a set of (proposed) values satisfying the following property: if the set of a non-faulty process is a singleton containing value v, the set of any non-faulty process contains v. Both communication abstractions have an O(n2) message complexity and a constant time complexity. The reduction of multivalued Byzantine consensus to binary Byzantine consensus is then a simple sequential use of these communication abstractions. To the best of our knowledge, this is the first asynchronous message-passing algorithm that reduces multivalued consensus to binary consensus with O(n2) messages and constant time complexity (measured with the longest causal chain of messages) in the presence of up to t<n/3t<n/3 Byzantine processes, and without using cryptography techniques. Moreover, this reduction algorithm uses a single instance of the underlying binary consensus, and tolerates message re-ordering by Byzantine processes. This work, done in collaboration with Achour Mostefaoui from LS2N (Nantes), appeared in Acta Informatica [20].

A distributed leader election algorithm in crash-recovery and omissive system

Participant : Michel Raynal.

We introduced a new distributed leader election algorithm for crash-recovery and omission environments. Contrary to previous works, our algorithm tolerates the occurrence of crash-recoveries and message omissions to any process during some finite but unknown time, after which a majority of processes in the system remains up and does not omit messages. This work, done in collaboration with Christian Fernández-Campusano, Mikel Larrea, and Roberto Cortiñas from UPV/EHU, Spain, appeared in Information Processing Letters 2017 [16].

Providing Collision-Free and Conflict-Free Communication in General Synchronous Broadcast/Receive Networks

Participants : Michel Raynal, François Taïani.

This work [26] considers the problem of communication in dense and large scale wireless networks composed of resource-limited nodes. In this kind of networks, a massive amount of data is becoming increasingly available, and consequently implementing protocols achieving error-free communication channels constitutes an important challenge. Indeed, in this kind of networks, the prevention of message conflicts and message collisions is a crucial issue. In terms of graph theory, solving this issue amounts to solve the distance-2 coloring problem in an arbitrary graph. The work presents a distributed algorithm providing the processes with such a coloring. This algorithm is itself collision-free and conflict-free. It is particularly suited to wireless networks composed of nodes with communication or local memory constraints.

This work was performed in collaboration with Abdelmadjid Bouabdallah and Hicham Lakhlef from Université Technologique de Compiègne (France).

Randomized abortable mutual exclusion with constant amortized RMR complexity on the CC model.

Participant : George Giakkoupis.

In [30], we presented an abortable mutual exclusion algorithm for the cache-coherent (CC) model with atomic registers and CAS objects. The algorithm has constant expected amortized RMR complexity in the oblivious adversary model and is deterministically deadlock-free. This is the first abortable mutual exclusion algorithm that achieves o(logn/loglogn) RMR complexity.

This work was done in collaboration with Philipp Woelfel (University of Calgary).