Presence Of Crashes Research Articles

Memento: A Framework for Detectable Recoverability in Persistent Memory

Persistent memory (PM) is an emerging class of storage technology that combines the performance of DRAM with the durability of SSD, offering the best of both worlds. This had led to a surge of research on persistent objects in PM. Among such persistent objects, concurrent data structures (DSs) are particularly interesting thanks to their performance and scalability. One of the most widely used correctness criteria for persistent concurrent DSs is detectable recoverability , ensuring both thread safety (for correctness in non-crashing concurrent executions) and crash consistency (for correctness in crashing executions). However, the existing approaches to designing detectably recoverable concurrent DSs are either limited to simple algorithms or suffer from high runtime overheads. We present Memento: a general and high-performance programming framework for detectably recoverable concurrent DSs in PM. To ensure general applicability to various DSs, Memento supports primitive operations such as checkpoint and compare-and-swap and their composition with control constructs. To ensure high performance, Memento employs a timestamp-based recovery strategy that requires fewer writes and flushes to PM than the existing approaches. We formally prove that Memento ensures detectable recoverability in the presence of crashes. To showcase Memento, we implement a lock-free stack, list, queue, and hash table, and a combining queue that detectably recovers from random crashes in stress tests and performs comparably to existing hand-tuned persistent DSs with and without detectable recoverability.

Research for Practice

Keeping data safe in the presence of crashes is a fundamental problem.

Crash Consistency

Keeping data safe in the presence of crashes is a fundamental problem in storage systems. Although the high-level ideas for crash consistency are relatively well understood, realizing them in practice is surprisingly complex and full of challenges. The systems research community is actively working on solving this challenge, and the papers examined here offer three solutions.

Awareness of Crash Risk Improves Kelly Strategies in Simulated Financial Time Series

We simulate a simplified version of the price process including bubbles and crashes proposed in Kreuser and Sornette (2018). The price process is defined as a geometric random walk combined with jumps modelled by separate, discrete distributions associated with positive (and negative) bubbles. The key ingredient of the model is to assume that the sizes of the jumps are proportional to the bubble size. Thus, the jumps tend to efficiently bring back excess bubble prices close to a “normal” or fundamental value (“efficient crashes”). This is different from existing processes studied that assume jumps that are independent of the mispricing. The present model is simplified compared to Kreuser and Sornette (2018) in that we ignore the possibility of a change of the probability of a crash as the price accelerates above the normal price. We study the behaviour of investment strategies that maximize the expected log of wealth (Kelly criterion) for the risky asset and a risk-free asset. We show that the method behaves similarly to Kelly on Geometric Brownian Motion in that it outperforms other methods in the long-term and it beats classical Kelly. As a primary source of outperformance, we determine knowledge about the presence of crashes, but interestingly find that knowledge of only the size, and not the time of occurrence, already provides a significant and robust edge. We then perform an error analysis to show that the method is robust with respect to variations in the parameters. The method is most sensitive to errors in the expected return.

Adaptive timeliness of consensus in presence of crash and timing faults

The Δ -timed uniform consensus is a stronger variant of the traditional consensus and it satisfies the following additional property: every correct process terminates its execution within a constant time Δ ( Δ -timeliness), and no two processes decide differently (uniformity). In this paper, we consider the Δ -timed uniform consensus problem in presence of f c crash processes and f t timing-faulty processes, and propose a Δ -timed uniform consensus algorithm. The proposed algorithm is adaptive in the following sense: it solves the Δ -timed uniform consensus when at least f t + 1 correct processes exist in the system. If the system has less than f t + 1 correct processes, the algorithm cannot solve the Δ -timed uniform consensus. However, as long as f t + 1 processes are non-crashed, the algorithm solves (non-timed) uniform consensus. We also investigate the maximum number of faulty processes that can be tolerated. We show that any Δ -timed uniform consensus algorithm tolerating up to f t timing-faulty processes requires that the system has at least f t + 1 correct processes. This impossibility result implies that the proposed algorithm attains the maximum resilience about the number of faulty processes. We also show that any Δ -timed uniform consensus algorithm tolerating up to f t timing-faulty processes cannot solve the (non-timed) uniform consensus when the system has less than f t + 1 non-crashed processes. This impossibility result implies that our algorithm attains the maximum adaptiveness.

Timed Uniform Atomic Broadcast in Presence of Crash and Timing Faults

SUMMARY ∆-Timed Atomic Broadcast is the broadcast ensuring that all correct processes deliver the same messages in the same order, and that delivery latency of any message broadcast by any correct process is some predetermined time ∆ or less. In this paper, we propose a ∆-timed atomic broadcast algorithm in a synchronous system where communication delay is bounded by a known constant d and processes suffer both crash faults and timing faults. The proposed algorithm can tolerate fc crash faults and ft timing faults as long as at least ft + 1 processes are correct, and its max