Consider an asynchronous network in a shared-memory environment consisting of n nodes. Assume that up to f of the nodes might be Byzantine (n > 12f), where the adversary is full-information and dynamic (sometimes called adaptive). In addition, the non-Byzantine nodes may undergo transient failures. Nodes advance in atomic steps, which consist of reading all registers, performing some calculation and writing to all registers.
This paper contains three main contributions. First, the clock-function problem is defined, which is a generalization of the clock synchronization problem. This generalization encapsulates previous clock synchronization problem definitions while extending them to the current paper's model. Second, a randomized asynchronous self-stabilizing Byzantine tolerant clock synchronization algorithm is presented.
In the construction of the clock synchronization algorithm, a building block that ensures different nodes advance at similar rates is developed. This feature is the third contribution of the paper. It is self-stabilizing and Byzantine tolerant and can be used as a building block for different algorithms that operate in an asynchronous self-stabilizing Byzantine model.
The convergence time of the presented algorithm is exponential. Observe that in the asynchronous setting the best known full-information dynamic Byzantine agreement also has expected exponential convergence time, even though currently there is no known reduction between the two.
In this paper, we propose a fully distributed algorithm for joint clock skew and offset estimation in wireless sensor networks based on belief propagation. In the proposed algorithm, each node can estimate its clock skew and offset in a completely distributed and asynchronous way: some nodes may update their estimates more frequently than others using outdated message from neighboring nodes. In addition, the proposed algorithm is robust to random packet loss. Such algorithm does not require any centralized information processing or coordination, and is scalable with network size. The proposed …
Consensus protocols for asynchronous networks are usually complex and inefficient, leading practical systems to rely on synchronous protocols. This paper attempts to simplify asynchronous consensus by building atop a novel threshold logical clock abstraction, which enables upper layers to operate as if on a synchronous network. This approach yields an asynchronous consensus protocol for fail-stop nodes that may be simpler and more robust than Paxos and its leader-based variants, requiring no common coins and achieving consensus in a constant expected number of rounds. The same approach can be …
It is commonly held that asynchronous consensus is much more complex, difficult, and costly than partially-synchronous algorithms, especially without using common coins. This paper challenges that conventional wisdom with que sera consensus QSC, an approach to consensus that cleanly decomposes the agreement problem from that of network asynchrony. QSC uses only private coins and reaches consensus in $O(1)$ expected communication rounds. It relies on "lock-step" synchronous broadcast, but can run atop a threshold logical clock (TLC) algorithm to time and pace partially-reliable communication at…
Phase clocks are synchronization tools that implement a form of logical time in distributed systems. For systems tolerating transient faults by self-repair of damaged data, phase clocks can enable reasoning about the progress of distributed repair procedures. This paper presents a phase clock algorithm suited to the model of transient memory faults in asynchronous systems with read/write registers. The algorithm is self-stabilizing and guarantees accuracy of phase clocks within O(k) time following an initial state that is k-faulty. Composition theorems show how the algorithm can be used for th…
Federated Edge Learning (FEL) allows edge nodes to train a global deep learning model collaboratively for edge computing in the Industrial Internet of Things (IIoT), which significantly promotes the development of Industrial 4.0. However, FEL faces two critical challenges: communication overhead and data privacy. FEL suffers from expensive communication overhead when training large-scale multi-node models. Furthermore, due to the vulnerability of FEL to gradient leakage and label-flipping attacks, the training process of the global model is easily compromised by adversaries. To address these c…
We study distributed plurality consensus among $n$ nodes, each of which initially holds one of $k$ opinions. The goal is to eventually agree on the initially dominant opinion. We consider an asynchronous communication model in which each node is equipped with a random clock. Whenever the clock of a node ticks, it may open communication channels to a constant number of other nodes, chosen uniformly at random or from a list of constantly many addresses acquired in previous steps. The tick rates and the delays for establishing communication channels (channel delays) follow some probability distri…
Today's hardware technology presents a new challenge in designing robust systems. Deep submicron VLSI technology introduced transient and permanent faults that were never considered in low-level system designs in the past. Still, robustness of that part of the system is crucial and needs to be guaranteed for any successful product. Distributed systems, on the other hand, have been dealing with similar issues for decades. However, neither the basic abstractions nor the complexity of contemporary fault-tolerant distributed algorithms match the peculiarities of hardware implementations. This pape…
Runtime verification is the process of verifying critical behavioral properties in big complex systems, where formal verification is not possible due to state space explosion. There have been several attempts to design efficient algorithms for runtime verification. Most of these algorithms have a formally defined correctness property as a reference and check whether the system consistently meets the demands of the property or it fails to satisfy the property at some point in runtime. LTL is a commonly used language for defining these kinds of properties and is also the language of focus in thi…
Vector clock algorithms are basic wait-free building blocks that facilitate causal ordering of events. As wait-free algorithms, they are guaranteed to complete their operations within a finite number of steps. Stabilizing algorithms allow the system to recover after the occurrence of transient faults, such as soft errors and arbitrary violations of the assumptions according to which the system was designed to behave. We present the first, to the best of our knowledge, stabilizing vector clock algorithm for asynchronous crash-prone message-passing systems that can recover in a wait-free manner …
Hong Kong University of Science and Technology（香港科技大学）
With the recent proliferation of large-scale learning problems,there have been a lot of interest on distributed machine learning algorithms, particularly those that are based on stochastic gradient descent (SGD) and its variants. However, existing algorithms either suffer from slow convergence due to the inherent variance of stochastic gradients, or have a fast linear convergence rate but at the expense of poorer solution quality. In this paper, we combine their merits by proposing a fast distributed asynchronous SGD-based algorithm with variance reduction. A constant learning rate can be used…
Recently, a detailed epistemic reasoning framework for multi-agent systems with byzantine faulty asynchronous agents and possibly unreliable communication was introduced. We have developed a modular extension framework implemented on top of it, which allows to encode and safely combine additional system assumptions commonly used in the modeling and analysis of fault-tolerant distributed systems, like reliable communication, time-bounded communication, multicasting, synchronous and lock-step synchronous agents and even agents with coordinated actions. We use this extension framework for analyzi…
We present concept and implementation of a self-stabilizing Byzantine fault-tolerant distributed clock generation scheme for multi-synchronous GALS architectures in critical applications. It combines a variant of a recently introduced self-stabilizing algorithm for generating low-frequency, low-accuracy synchronized pulses with a simple non-stabilizing high-frequency, high-accuracy clock synchronization algorithm. We provide thorough correctness proofs and a performance analysis, which use methods from fault-tolerant distributed computing research but also addresses hardware-related issues lik…
Propelled by the growth of large-scale blockchain deployments, much recent progress has been made in designing sharding protocols that achieve throughput scaling linearly in the number of nodes. However, existing protocols are not robust to an adversary adaptively corrupting a fixed fraction of nodes. In this paper, we propose Free2Shard -- a new architecture that achieves near-linear scaling while being secure against a fully adaptive adversary. The focal point of this architecture is a dynamic self-allocation algorithm that lets users allocate themselves to shards in response to adversarial …
We consider distributed plurality consensus in a complete graph of size $n$ with $k$ initial opinions. We design an efficient and simple protocol in the asynchronous communication model that ensures that all nodes eventually agree on the initially most frequent opinion. In this model, each node is equipped with a random Poisson clock with parameter $λ=1$. Whenever a node's clock ticks, it samples some neighbors, uniformly at random and with replacement, and adjusts its opinion according to the sample. A prominent example is the so-called two-choices algorithm in the synchronous model, where in…
University of Southern California（南加利福尼亚大学）；University of Illinois at Urbana-Champaign（伊利诺伊大学厄巴纳-香槟分校）
Distributed training of deep nets is an important technique to address some of the present day computing challenges like memory consumption and computational demands. Classical distributed approaches, synchronous or asynchronous, are based on the parameter server architecture, i.e., worker nodes compute gradients which are communicated to the parameter server while updated parameters are returned. Recently, distributed training with AllReduce operations gained popularity as well. While many of those operations seem appealing, little is reported about wall-clock training time improvements. In t…
Sandia National Laboratories（桑迪亚国家实验室）；Xilinx Inc.（赛灵思公司）
Bayesian optimization (BO) is a flexible and powerful framework that is suitable for computationally expensive simulation-based applications and guarantees statistical convergence to the global optimum. While remaining as one of the most popular optimization methods, its capability is hindered by the size of data, the dimensionality of the considered problem, and the nature of sequential optimization. These scalability issues are intertwined with each other and must be tackled simultaneously. In this work, we propose the Scalable$^3$-BO framework, which employs sparse GP as the underlying surr…
We propose an asynchronous, decentralized algorithm for consensus optimization. The algorithm runs over a network in which the agents communicate with their neighbors and perform local computation. In the proposed algorithm, each agent can compute and communicate independently at different times, for different durations, with the information it has even if the latest information from its neighbors is not yet available. Such an asynchronous algorithm reduces the time that agents would otherwise waste idle because of communication delays or because their neighbors are slower. It also eliminates …
