arXiv subjects

Michel Raynal

Publications and source records attributed to Michel Raynal.

At least 37 records · Page 2Linked to original sources

arXiv2021

Self-stabilizing Multivalued Consensus in Asynchronous Crash-prone Systems

The problem of multivalued consensus is fundamental in the area of fault-tolerant distributed computing since it abstracts a very broad set of agreement problems in which processes have to uniformly decide on a specific value v in V, where |V| >1. Existing solutions (that tolerate process failures) reduce the multivalued consensus problem to the one of binary consensus, e.g., Mostefaoui-Raynal-Tronel and Zhang-Chen. Our study aims at the design of an even more reliable solution. We do so through the lenses of self-stabilization -- a very strong notion of fault-tolerance. In addition to node and communication failures, self-stabilizing algorithms can recover after the occurrence of arbitrary transient-faults; these faults represent any violation of the assumptions according to which the system was designed to operate (as long as the algorithm code stays intact). This work proposes the first (to the best of our knowledge) self-stabilizing algorithm for multivalued consensus for asynchronous message-passing systems prone to process failures and arbitrary transient-faults. Our solution is also the first (to the best of our knowledge) to support wait-freedom. Moreover, using piggybacking techniques, our solution can invoke n binary consensus objects concurrently. Thus, the proposed self-stabilizing wait-free solution can terminate using fewer resources than earlier non-self-stabilizing solutions by Mostefaoui, Raynal, and Tronel, which uses an unbounded number of binary consensus objects, or Zhang and Chen, which is not wait-free.

cs.DC↗

arXiv2021

Byzantine-tolerant Distributed Grow-only Sets: Specification and Applications

In order to formalize Distributed Ledger Technologies and their interconnections, a recent line of research work has formulated the notion of Distributed Ledger Object (DLO), which is a concurrent object that maintains a totally ordered sequence of records, abstracting blockchains and distributed ledgers. Through DLO, the Atomic Appends problem, intended as the need of a primitive able to append multiple records to distinct ledgers in an atomic way, is studied as a basic interconnection problem among ledgers. In this work, we propose the Distributed Grow-only Set object (DSO), which instead of maintaining a sequence of records, as in a DLO, maintains a set of records in an immutable way: only Add and Get operations are provided. This object is inspired by the Grow-only Set (G-Set) data type which is part of the Conflict-free Replicated Data Types. We formally specify the object and we provide a consensus-free Byzantine-tolerant implementation that guarantees eventual consistency. We then use our Byzantine-tolerant DSO (BDSO) implementation to provide consensus-free algorithmic solutions to the Atomic Appends and Atomic Adds (the analogous problem of atomic appends applied on G-Sets) problems, as well as to construct consensus-free Single-Writer BDLOs. We believe that the BDSO has applications beyond the above-mentioned problems.

cs.DC↗

arXiv2021

Money Transfer Made Simple: a Specification, a Generic Algorithm, and its Proof

It has recently been shown that, contrarily to a common belief, money transfer in the presence of faulty (Byzantine) processes does not require strong agreement such as consensus. This article goes one step further: namely, it first proposes a non-sequential specification of the money-transfer object, and then presents a generic algorithm based on a simple FIFO order between each pair of processes that implements it. The genericity dimension lies in the underlying reliable broadcast abstraction which must be suited to the appropriate failure model. Interestingly, whatever the failure model, the money transfer algorithm only requires adding a single sequence number to its messages as control information. Moreover, as a side effect of the proposed algorithm, it follows that money transfer is a weaker problem than the construction of a safe/regular/atomic read/write register in the asynchronous message-passing crash-prone model.

cs.DC↗

arXiv2020

Relaxed Queues and Stacks from Read/Write Operations

Considering asynchronous shared memory systems in which any number of processes may crash, this work identifies and formally defines relaxations of queues and stacks that can be non-blocking or wait-free while being implemented using only read/write operations. Set-linearizability and Interval-linearizability are used to specify the relaxations formally, and precisely identify the subset of executions which preserve the original sequential behavior. The relaxations allow for an item to be returned more than once by different operations, but only in case of concurrency; we call such a property multiplicity. The stack implementation is wait-free, while the queue implementation is non-blocking. Interval-linearizability is used to describe a queue with multiplicity, with the additional relaxation that a dequeue operation can return weak-empty, which means that the queue might be empty. We present a read/write wait-free interval-linearizable algorithm of a concurrent queue. As far as we know, this work is the first that provides formalizations of the notions of multiplicity and weak-emptiness, which can be implemented on top of read/write registers only.

cs.DC↗

arXiv2020

Self-Stabilizing Indulgent Zero-degrading Binary Consensus

Guerraoui proposed an indulgent solution for the binary consensus problem. Namely, he showed that an arbitrary behavior of the failure detector never violates safety requirements even if it compromises liveness. Consensus implementations are often used in a repeated manner. Dutta and Guerraoui proposed a zero-degrading solution, \ie during system runs in which the failure detector behaves perfectly, a node failure during one consensus instance has no impact on the performance of future instances. Our study, which focuses on indulgent zero-degrading binary consensus, aims at the design of an even more robust communication abstraction. We do so through the lenses of self-stabilization - a very strong notion of fault-tolerance. In addition to node and communication failures, self-stabilizing algorithms can recover after the occurrence of arbitrary transient faults; these faults represent any violation of the assumptions according to which the system was designed to operate (as long as the algorithm code stays intact). This work proposes the first, to the best of our knowledge, self-stabilizing algorithm for indulgent zero-degrading binary consensus for time-free message-passing systems prone to detectable process failures. The proposed algorithm has an O(1) stabilization time (in terms of asynchronous cycles) from arbitrary transient faults. Since the proposed solution uses an Ω failure detector, we also present the first, to the best of our knowledge, self-stabilizing asynchronous Ω failure detector, which is a variation on the one by Mostéfaoui, Mourgaya, and Raynal.

cs.DC↗

arXiv2020

$t$-Resilient $k$-Immediate Snapshot and its Relation with Agreement Problems

An immediate snapshot object is a high level communication object, built on top of a read/write distributed system in which all except one processes may crash. It allows a process to write a value and obtain a set of values that represent a snapshot of the values written to the object, occurring immediately after the write step. Considering an $n$-process model in which up to $t$ processes may crash, this paper introduces first the $k$-resilient immediate snapshot object, which is a natural generalization of the basic immediate snapshot (which corresponds to the case $k=t=n-1$). In addition to the set containment properties of the basic immediate snapshot, a $k$-resilient immediate snapshot object requires that each set returned to a process contains at least $(n-k)$ pairs. The paper first shows that, for $k,t<n-1$, $k$-resilient immediate snapshot is impossible in asynchronous read/write systems. %Then the paper investigates the space of objects that %are impossible to solve in $n$-process $t$-crash read/write systems. Then the paper investigates a model of computation where the processes communicate with each other by accessing $k$-immediate snapshot objects, and shows that this model is stronger than the $t$-crash model. Considering the space of $x$-set agreement problems (which are impossible to solve in systems such that $x\leq t$), the paper shows then that $x$-set agreement can be solved in read/write systems enriched with $k$-immediate snapshot objects for $x=\max(1,t+k-(n-2))$. It also shows that, in these systems, $k$-resilient immediate snapshot and consensus are equivalent when $1\leq t<n/2$ and $t\leq k\leq (n-1)-t$. Hence, %thanks to the problem map it provides, the paper establishes strong relations linking fundamental distributed computing objects (one related to communication, the other to agreement), which are impossible to solve in pure read/write systems.

cs.DC↗

arXiv2020

Appending Atomically in Byzantine Distributed Ledgers

A Distributed Ledger Object (DLO) is a concurrent object that maintains a totally ordered sequence of records, and supports two basic operations: append, which appends a record at the end of the sequence, and get, which returns the sequence of records. In this work we provide a proper formalization of a Byzantine-tolerant Distributed Ledger Object (BDLO), which is a DLO in a distributed system in which processes may deviate arbitrarily from their indented behavior, i.e. they may be Byzantine. Our formal definition is accompanied by algorithms to implement BDLOs by utilizing an underlying Byzantine Atomic Broadcast service. We then utilize the BDLO implementations to solve the Atomic Appends problem against Byzantine processes. The Atomic Appends problem emerges when several clients have records to append, the record of each client has to be appended to a different BDLO, and it must be guaranteed that either all records are appended or none. We present distributed algorithms implementing solutions for the Atomic Appends problem when the clients (which are involved in the appends) and the servers (which maintain the BDLOs) may be Byzantine.

cs.DC↗

arXiv2020

Self-stabilizing Uniform Reliable Broadcast

We study a well-known communication abstraction called Uniform Reliable Broadcast (URB). URB is central in the design and implementation of fault-tolerant distributed systems, as many non-trivial fault-tolerant distributed applications require communication with provable guarantees on message deliveries. Our study focuses on fault-tolerant implementations for time-free message-passing systems that are prone to node-failures. Moreover, we aim at the design of an even more robust communication abstraction. We do so through the lenses of self-stabilization---a very strong notion of fault-tolerance. In addition to node and communication failures, self-stabilizing algorithms can recover after the occurrence of arbitrary transient faults; these faults represent any violation of the assumptions according to which the system was designed to operate (as long as the algorithm code stays intact). This work proposes the first self-stabilizing URB solution for time-free message-passing systems that are prone to node-failures. The proposed algorithm has an O(bufferUnitSize) stabilization time (in terms of asynchronous cycles) from arbitrary transient faults, where bufferUnitSize is a predefined constant that can be set according to the available memory. Moreover, the communication costs of our algorithm are similar to the ones of the non-self-stabilizing state-of-the-art. The main differences are that our proposal considers repeated gossiping of O(1) bits messages and deals with bounded space (which is a prerequisite for self-stabilization). Specifically, each node needs to store up to bufferUnitSize n records and each record is of size O(v + n log n) bits, where n is the number of nodes in the system and v is the number of bits needed to encode a single URB instance.

cs.DC↗

arXiv2019

Fully Anonymous Shared Memory Algorithms

Process anonymity has been studied for a long time. Memory anonymity is more recent. In an anonymous memory system, there is no a priori agreement among the processes on the names of the shared registers they access. This article introduces the fully anonymous model, namely a model in which both the processes and the memory are anonymous. It is shown that fundamental problems such as mutual exclusion, consensus, and its weak version called set agreement, can be solved despite full anonymity, the first in a failure-free system, the others in the presence of any number of process crashes.

cs.DC↗

arXiv2019

Mutex-based Desanonymization of an Anonymous Read/Write Memory

Anonymous shared memory is a memory in which processes use different names for the same shared read/write register. As an example, a shared register named $A$ by a process $p$ and a shared register named $B$ by another process $q$ can correspond to the very same register $X$, and similarly for the names $B$ at $p$ and $A$ at $q$ which can correspond to the same register $Y\neq X$. Hence, there is a permanent disagreement on the register names among the processes. This new notion of anonymity was recently introduced by G. Taubenfeld (PODC 2017), who presented several memory-anonymous algorithms and impossibility results. This paper introduces a new problem (new to our knowledge), that consists in "desanonymizing" an anonymous shared memory. To this end, it presents an algorithm that, starting with a shared memory made up of $m$ anonymous read/write atomic registers (i.e., there is no a priori agreement on their names), allows each process to compute a local addressing mapping, such that all the processes agree on the names of each register. The proposed construction is based on an underlying deadlock-free mutex algorithm for $n\geq 2$ processes (recently proposed in a paper co-authored by some of the authors of this paper), and consequently inherits its necessary and sufficient condition on the size $m$ of the anonymous memory, namely $m$ must belongs to the set $M(n)=\{m:~$ such that $\forall~ \ell: 1<\ell \leq n:~ \gcd(\ell,m)=1\}\setminus \{1\}$. This algorithm, which is also symmetric in the sense process identities can only be compared by equality, requires the participation of all the processes; hence it can be part of the system initialization. Last but not least, the proposed algorithm has a first-class noteworthy property, namely, its simplicity.

cs.DC↗

arXiv2018

Mastering Concurrent Computing Through Sequential Thinking: A Half-century Evolution

Concurrency, the art of doing many things at the same time is slowly becoming a science. It is very difficult to master, yet it arises all over modern computing systems, both when the communication medium is shared memory and when it is by message passing. Concurrent programming is hard because it requires to cope with many possible, unpredictable behaviors of communicating processes interacting with each other. Right from the start in the 1960s, the main way of dealing with concurrency has been by reduction to sequential reasoning. We trace this history, and illustrate it through several examples, from early ideas based on mutual exclusion, passing through consensus and concurrent objects, until today ledgers and blockchains. We conclude with a discussion on the limits that this approach encounters, related to fault-tolerance, performance, and inherently concurrent problems.

cs.DC↗

arXiv2018

Optimal Memory-Anonymous Symmetric Deadlock-Free Mutual Exclusion

The notion of an anonymous shared memory (recently introduced in PODC 2017) considers that processes use different names for the same memory location. Hence, there is permanent disagreement on the location names among processes. In this context, the PODC paper presented -among other results- a symmetric deadlock-free mutual exclusion (mutex) algorithm for two processes and a necessary condition on the size $m$ of the anonymous memory for the existence of a symmetric deadlock-free mutex algorithm in an $n$-process system. This condition states that $m$ must be greater than $1$ and belong to the set $M(n)=\{m:\forall~\ell:1<\ell\leq n:~\gcd(\ell,m)=1\}$ (symmetric means that, while each process has its own identity, process identities can only be compared with equality). The present paper answers several open problems related to symmetric deadlock-free mutual exclusion in an $n$-process system ($n\geq 2$) where the processes communicate through $m$ registers. It first presents two algorithms. The first considers that the registers are anonymous read/write atomic registers and works for any $m$ greater than $1$ and belonging to the set $M(n)$. It thus shows that this condition on $m$ is both necessary and sufficient. The second algorithm considers anonymous read/modify/write atomic registers. It assumes that $m\in M(n)$. These algorithms differ in their design principles and their costs (measured as the number of registers which must contain the identity of a process to allow it to enter the critical section). The paper also shows that the condition $m\in M(n)$ is necessary for deadlock-free mutex on top of anonymous read/modify/write atomic registers. It follows that, when $m>1$, $m\in M(n)$ is a tight characterization of the size of the anonymous shared memory needed to solve deadlock-free mutex, be the anonymous registers read/write or read/modify/write.

cs.DC↗

arXiv2018

DBFT: Efficient Byzantine Consensus with a Weak Coordinator and its Application to Consortium Blockchains

This paper introduces a deterministic Byzantine consensus algorithm that relies on a new weak coordinator. As opposed to previous algorithms that cannot terminate in the presence of a faulty or slow coordinator, our algorithm can terminate even when its coordinator is faulty, hence the name weak coordinator. The key idea is to allow processes to complete asynchronous rounds as soon as they receive a threshold of messages, instead of having to wait for a message from a coordinator that may be slow. The resulting algorithm assumes partial synchrony, is resilience optimal, time optimal and does not need signatures. Our presentation is didactic: we first present a simple safe binary Byzantine consensus algorithm, modify it to ensure termination, and finally present an optimized reduction from multivalue consensus to binary consensus that may terminate in 4 message delays. To evaluate our algorithm, we deployed it on 100 machines distributed in 5 datacenters across different continents and compared its performance against the randomized solution from Mostefaoui, Moumem and Raynal [PODC14] that terminates in O(1) rounds in expectation. Our algorithm always outperforms the latter even in the presence of Byzantine behaviors. Our algorithm has a subsecond average latency in most of our geo-distributed experiments, even when attacked by a well-engineered coalition of Byzantine processes.

cs.DC↗

arXiv2018

A Simple Object that Spans the Whole Consensus Hierarchy

This paper presents a simple generalization of the basic atomic read/write register object, whose genericity parameter spans the whole set of integers and is such that its k-parameterized instance has exactly consensus number k. This object, whose definition is pretty natural, is a sliding window register of size k. Its interest lies in its simplicity and its genericity dimension which provides a global view capturing the whole consensus hierarchy. Hence, this short article must be seen as a simple pedagogical introduction to Herlihy's consensus hierarchy.

cs.DC↗

arXiv2018

Extending Causal Consistency to any Object Defined by a Sequential Specification

This paper presents a simple generalization of causal consistency suited to any object defined by a sequential specification. As causality is captured by a partial order on the set of operations issued by the processes on shared objects (concurrent operations are not ordered), it follows that causal consistency allows different processes to have different views of each object history.

cs.DC↗

arXiv2017

Set-Constrained Delivery Broadcast: Definition, Abstraction Power, and Computability Limits

This paper introduces a new communication abstraction, called Set-Constrained Delivery Broadcast (SCD-broadcast), whose aim is to provide its users with an appropriate abstraction level when they have to implement objects or distributed tasks in an asynchronous message-passing system prone to process crash failures. This abstraction allows each process to broadcast messages and deliver a sequence of sets of messages in such a way that, if a process delivers a set of messages including a message m and later delivers a set of messages including a message m ' , no process delivers first a set of messages including m ' and later a set of message including m. After having presented an algorithm implementing SCD-broadcast, the paper investigates its programming power and its computability limits. On the "power" side it presents SCD-broadcast-based algorithms, which are both simple and efficient, building objects (such as snapshot and conflict-free replicated data), and distributed tasks. On the "computability limits" side it shows that SCD-broadcast and read/write registers are computationally equivalent.

cs.DC↗

arXiv2017

Which Broadcast Abstraction Captures $k$-Set Agreement?

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, this article addresses and answers the following question: which is the communication abstraction that "captures" $k$-set agreement? To this end, it introduces 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.

cs.DC↗

arXiv2017

Another Look at the Implementation of Read/write Registers in Crash-prone Asynchronous Message-Passing Systems (Extended Version)

" Yet another paper on " the implementation of read/write registers in crash-prone asynchronous message-passing systems! Yes..., but, differently from its predecessors, this paper looks for a communication abstraction which captures the essence of such an implementation in the same sense that total order broadcast can be associated with consensus, or message causal delivery can be associated with causal read/write registers. To this end, the paper introduces a new communication abstraction, named SCD-broadcast (SCD standing for " Set Constrained Delivery "), which, instead of a single message, delivers to processes sets of messages (whose size can be arbitrary), such that the sequences of message sets delivered to any two processes satisfies some constraints. The paper then shows that: (a) SCD-broadcast allows for a very simple implementation of a snapshot object (and consequently also of atomic read/write registers) in crash-prone asynchronous message-passing systems, (b) SCD-broadcast can be built from snapshot objects (hence SCD-broadcast and snapshot objects --or read/write registers-- are " computationally equivalent "), (c) SCD-broadcast can be built in message-passing systems where any minority of processes may crash (which is the weakest assumption on the number of possible process crashes needed to implement a read/write register).

cs.DC↗