We present an efficient lock-free algorithm for parallel accessible hash tables with open addressing, which promises more robust performance and reliability than conventional lock-based implementations. ``Lock-free'' means that it is guaranteed that always at least one process completes its operation within a bounded number of steps. For a single processor architecture our solution is as efficient as sequential hash tables. On a multiprocessor architecture this is also the case when all processors have comparable speeds. The algorithm allows processors that have widely different speeds or come to a halt. It can easily be implemented using C-like languages and requires on average only constant time for insertion, deletion or accessing of elements. The algorithm allows the hash tables to grow and shrink when needed.
Lock-free algorithms are hard to design correctly, even when apparently straightforward. Ensuring the correctness of the design at the earliest possible stage is a major challenge in any responsible system development. In view of the complexity of the algorithm, we turned to the interactive theorem prover PVS for mechanical support. We employ standard deductive verification techniques to prove around 200 invariance properties of our algorithm, and describe how this is achieved with the theorem prover PVS.
In this paper we present a lock-free version of Hopscotch Hashing. Hopscotch Hashing is an open addressing algorithm originally proposed by Herlihy, Shavit, and Tzafrir, which is known for fast performance and excellent cache locality. The algorithm allows users of the table to skip or jump over irrelevant entries, allowing quick search, insertion, and removal of entries. Unlike traditional linear probing, Hopscotch Hashing is capable of operating under a high load factor, as probe counts remain small. Our lock-free version improves on both speed, cache locality, and progress guarantees of the…
In the past few years, General Purpose Graphics Processors (GPUs) have been used to significantly speed up numerous applications. One of the areas in which GPUs have recently led to a significant speed-up is model checking. In model checking, state spaces, i.e., large directed graphs, are explored to verify whether models satisfy desirable properties. GPUexplore is a GPU-based model checker that uses a hash table to efficiently keep track of already explored states. As a large number of states is discovered and stored during such an exploration, the hash table should be able to quickly handle …
Hash tables are used in a plethora of applications, including database operations, DNA sequencing, string searching, and many more. As such, there are many parallelized hash tables targeting multicore, distributed, and accelerator-based systems. We present in this work a multi-GPU hash table implementation that can process keys at a throughput comparable to that of distributed hash tables. Distributed CPU hash tables have received significantly more attention than GPU-based hash tables. We show that a single node with multiple GPUs offers roughly the same performance as a 500-1,000-core CPU-ba…
Hash tables are ubiquitous and used in a wide range of applications for efficient probing of large and unsorted data. If designed properly, hash-tables can enable efficients look ups in a constant number of operations or commonly referred to as O(1) operations. As data sizes continue to grow and data becomes less structured (as is common for big-data applications), the need for efficient and scalable hash table also grows. In this paper we introduce HashGraph, a new scalable approach for building hash tables that uses concepts taken from sparse graph representations--hence the name HashGraph. …
Most cloud services and distributed applications rely on hashing algorithms that allow dynamic scaling of a robust and efficient hash table. Examples include AWS, Google Cloud and BitTorrent. Consistent and rendezvous hashing are algorithms that minimize key remapping as the hash table resizes. While memory errors in large-scale cloud deployments are common, neither algorithm offers both efficiency and robustness. Hyperdimensional Computing is an emerging computational model that has inherent efficiency, robustness and is well suited for vector or hardware acceleration. We propose Hyperdimensi…
This paper presents an efficient wait-free resizable hash table. To achieve high throughput at large core counts, our algorithm is specifically designed to retain the natural parallelism of concurrent hashing, while providing wait-free resizing. An extensive evaluation of our hash table shows that in the common case where resizing actions are rare, our implementation outperforms all existing lock-free hash table implementations while providing a stronger progress guarantee.
Given a specified average load factor, hash tables offer the appeal of constant time lookup operations. However, hash tables could face severe hash collisions because of malicious attacks, buggy applications, or even bursts of incoming data, compromising this practical advantage. In this paper, we present DHash, a hash table that overcomes this challenge by allowing programmers to dynamically change its hash function on the fly, without affecting other concurrent operations such as lookup, insert, and delete. DHash is modular and allows programmers to select a variety of lock-free/wait-free se…
Cuckoo hashing guarantees constant-time lookups regardless of table density, making it a viable candidate for high-density tables. Cuckoo hashing insertions perform poorly at high table densities, however. In this paper, we mitigate this problem through the introduction of novel kick-out eviction algorithms. Experimentally, our algorithms reduce the number of bins viewed per insertion for high-density tables by as much as a factor of ten. We also introduce an optimistic concurrency scheme for transactional multi-writer cuckoo hash tables (not using hardware transactional memory). For delete-li…
Hash table is a fundamental data structure for quick search and retrieval of data. It is a key component in complex graph analytics and AI/ML applications. State-of-the-art parallel hash table implementations either make some simplifying assumptions such as supporting only a subset of hash table operations or employ optimizations that lead to performance that is highly data dependent and in the worst case can be similar to a sequential implementation. In contrast, in this work we develop a dynamic hash table that supports all the hash table queries - search, insert, delete, update, while allow…
We design and implement a fully concurrent dynamic hash table for GPUs with comparable performance to the state of the art static hash tables. We propose a warp-cooperative work sharing strategy that reduces branch divergence and provides an efficient alternative to the traditional way of per-thread (or per-warp) work assignment and processing. By using this strategy, we build a dynamic non-blocking concurrent linked list, the slab list, that supports asynchronous, concurrent updates (insertions and deletions) as well as search queries. We use the slab list to implement a dynamic hash table wi…
This paper focuses on data structures for multi-core reachability, which is a key component in model checking algorithms and other verification methods. A cornerstone of an efficient solution is the storage of visited states. In related work, static partitioning of the state space was combined with thread-local storage and resulted in reasonable speedups, but left open whether improvements are possible. In this paper, we present a scaling solution for shared state storage which is based on a lockless hash table implementation. The solution is specifically designed for the cache architecture of…
We present a binary routing tree protocol for distributed hash table overlays. Using this protocol each peer can independently route messages to its parent and two descendants on the fly without any maintenance, global context, and synchronization. The protocol is then extended to support tree change notification with similar efficiency. The resulting tree is almost perfectly dense and balanced, and has O(1) stretch if the distributed hash table is symmetric Chord. We use the tree routing protocol to overcome the main impediment for implementation of local thresholding algorithms in peer-to-pe…
We propose a simple distributed hash table called ReCord, which is a generalized version of Randomized-Chord and offers improved tradeoffs in performance and topology maintenance over existing P2P systems. ReCord is scalable and can be easily implemented as an overlay network, and offers a good tradeoff between the node degree and query latency. For instance, an $n$-node ReCord with $O(\log n)$ node degree has an expected latency of $Θ(\log n)$ hops. Alternatively, it can also offer $Θ(\frac{\log n}{\log \log n})$ hops latency at a higher cost of $O(\frac{\log^2 n}{\log \log n})$ node degree. …
Hash tables are one of the most fundamental data structures for effectively storing and accessing sparse data, with widespread usage in domains ranging from computer graphics to machine learning. This study surveys the state-of-the-art research on data-parallel hashing techniques for emerging massively-parallel, many-core GPU architectures. Key factors affecting the performance of different hashing schemes are discovered and used to suggest best practices and pinpoint areas for further research.
Hash tables are ubiquitous. Properties such as an amortized constant time complexity for insertion and querying as well as a compact memory layout make them versatile associative data structures with manifold applications. The rapidly growing amount of data emerging in many fields motivated the need for accelerated hash tables designed for modern parallel architectures. In this work, we exploit the fast memory interface of modern GPUs together with a parallel hashing scheme tailored to improve global memory access patterns, to design WarpCore -- a versatile library of hash table data structure…
Distributed Hash Tables offer a resilient lookup service for unstable distributed environments. Resilient data storage, however, requires additional data replication and maintenance algorithms. These algorithms can have an impact on both the performance and the scalability of the system. In this paper, we describe the goals and design space of these replication algorithms. We examine an existing replication algorithm, and present a new analysis of its reliability. We then present a new dynamic replication algorithm which can operate in unstable environments. We give several possible replica pl…
Distributed hash table (DHT) is the foundation of many widely used storage systems, for its prominent features of high scalability and load balancing. Recently, DHT-based systems have been deployed for the Internet-of-Things (IoT) application scenarios. Unfortunately, such systems can experience a breakdown in the scale-out and load rebalancing process. This phenomenon contradicts with the common conception of DHT systems, especially about its scalability and load balancing features. In this paper, we investigate the breakdown of DHT-based systems in the scale-out process. We formulate the loa…
The first obstacle that regular users encounter when setting up a node for a public blockchain is the time taken for downloading all the data needed for the node to start operating correctly. In fact, this may last from hours to weeks for the major networks. Our contribution is twofold. Firstly, we show a design that enables mining and validation of new blocks keeping only a very small state. Secondly, we show that it is possible to store the state of the blockchain in a distributed hash table obtaining a wide spectrum of trade-offs between storage committed by the nodes and replication factor…
Many decentralized online social networks (DOSNs) have been proposed due to an increase in awareness related to privacy and scalability issues in centralized social networks. Such decentralized networks transfer processing and storage functionalities from the service providers towards the end users. DOSNs require individualistic implementation for services, (i.e., search, information dissemination, storage, and publish/subscribe). However, many of these services mostly perform social queries, where OSN users are interested in accessing information of their friends. In our work, we design a soc…
We propose a new and easily-realizable distributed hash table (DHT) peer-to-peer structure, incorporating a random caching strategy that allows for {\em polylogarithmic search time} while having only a {\em constant cache} size. We also show that a very large class of deterministic caching strategies, which covers almost all previously proposed DHT systems, can not achieve polylog search time with constant cache size. In general, the new scheme is the first known DHT structure with the following highly-desired properties: (a) Random caching strategy with constant cache size; (b) Average search…
A standard design pattern found in many concurrent data structures, such as hash tables or ordered containers, is alternation of parallelizable sections that incur no data conflicts and critical sections that must run sequentially and are protected with locks. It was already shown that simple stochastic analysis can predict the throughput of coarse-grained lock-based algorithms using CLH lock. In this short paper, we extend this analysis to algorithms based on the popular MCS lock.
Distributed Hash Tables (DHTs) have been used in several applications, but most DHTs have opted to solve lookups with multiple hops, to minimize bandwidth costs while sacrificing lookup latency. This paper presents D1HT, an original DHT which has a peer-to-peer and self-organizing architecture and maximizes lookup performance with reasonable maintenance traffic, and a Quarantine mechanism to reduce overheads caused by volatile peers. We implemented both D1HT and a prominent single-hop DHT, and we performed an extensive and highly representative DHT experimental comparison, followed by compleme…
Non-volatile memory is expected to co-exist or replace DRAM in upcoming architectures. Durable concurrent data structures for non-volatile memories are essential building blocks for constructing adequate software for use with these architectures. In this paper, we propose a new approach for durable concurrent sets and use this approach to build the most efficient durable hash tables available today. Evaluation shows a performance improvement factor of up to 3.3x over existing technology.
A key characteristic of deep recommendation models is the immense memory requirements of their embedding tables. These embedding tables can often reach hundreds of gigabytes which increases hardware requirements and training cost. A common technique to reduce model size is to hash all of the categorical variable identifiers (ids) into a smaller space. This hashing reduces the number of unique representations that must be stored in the embedding table; thus decreasing its size. However, this approach introduces collisions between semantically dissimilar ids that degrade model quality. We introd…
We present Recipe, a principled approach for converting concurrent DRAM indexes into crash-consistent indexes for persistent memory (PM). The main insight behind Recipe is that isolation provided by a certain class of concurrent in-memory indexes can be translated with small changes to crash-consistency when the same index is used in PM. We present a set of conditions that enable the identification of this class of DRAM indexes, and the actions to be taken to convert each index to be persistent. Based on these conditions and conversion actions, we modify five different DRAM indexes based on B+…
Mining frequent itemsets from massive datasets is always being a most important problem of data mining. Apriori is the most popular and simplest algorithm for frequent itemset mining. To enhance the efficiency and scalability of Apriori, a number of algorithms have been proposed addressing the design of efficient data structures, minimizing database scan and parallel and distributed processing. MapReduce is the emerging parallel and distributed technology to process big datasets on Hadoop Cluster. To mine big datasets it is essential to re-design the data mining algorithm on this new paradigm.…
We develop a family of parallel algorithms for the SpKAdd operation that adds a collection of k sparse matrices. SpKAdd is a much needed operation in many applications including distributed memory sparse matrix-matrix multiplication (SpGEMM), streaming accumulations of graphs, and algorithmic sparsification of the gradient updates in deep learning. While adding two sparse matrices is a common operation in Matlab, Python, Intel MKL, and various GraphBLAS libraries, these implementations do not perform well when adding a large collection of sparse matrices. We develop a series of algorithms usin…
In this paper we examine the issues involved in adding concurrency to the Robin Hood hash table algorithm. We present a non-blocking obstruction-free K-CAS Robin Hood algorithm which requires only a single word compare-and-swap primitive, thus making it highly portable. The implementation maintains the attractive properties of the original Robin Hood structure, such as a low expected probe length, capability to operate effectively under a high load factor and good cache locality, all of which are essential for high performance on modern computer architectures. We compare our data-structures to…
Department of Computer Science, University of Southern California（南加州大学计算机科学系）；Department of Computer Science and Technology, Tsinghua University（清华大学计算机科学与技术系）
We present a GPU-based Locality Sensitive Hashing (LSH) algorithm to speed up beam search for sequence models. We utilize the winner-take-all (WTA) hash, which is based on relative ranking order of hidden dimensions and thus resilient to perturbations in numerical values. Our algorithm is designed by fully considering the underling architecture of CUDA-enabled GPUs (Algorithm/Architecture Co-design): 1) A parallel Cuckoo hash table is applied for LSH code lookup (guaranteed O(1) lookup time); 2) Candidate lists are shared across beams to maximize the parallelism; 3) Top frequent words are merg…
