skip to main content


Title: Linear Probing Revisited: Tombstones Mark the Demise of Primary Clustering
First introduced in 1954, the linear-probing hash table is among the oldest data structures in computer science, and thanks to its unrivaled data locality, linear probing continues to be one of the fastest hash tables in practice. It is widely believed and taught, however, that linear probing should never be used at high load factors; this is because of an effect known as primary clustering which causes insertions at a load factor of 1 - 1/x to take expected time O(x2) (rather than the intuitive running time of ⇥(x)). The dangers of primary clustering, first discovered by Knuth in 1963, have now been taught to generations of computer scientists, and have influenced the design of some of the most widely used hash tables in production. We show that primary clustering is not the foregone conclusion that it is reputed to be. We demonstrate that seemingly small design decisions in how deletions are implemented have dramatic effects on the asymptotic performance of insertions: if these design decisions are made correctly, then even if a hash table operates continuously at a load factor of 1 - (1/x), the expected amortized cost per insertion/deletion is O(x). This is because the tombstones left behind by deletions can actually cause an anti-clustering effect that combats primary clustering. Interestingly, these design decisions, despite their remarkable effects, have historically been viewed as simply implementation-level engineering choices. We also present a new variant of linear probing (which we call graveyard hashing) that completely eliminates primary clustering on any sequence of operations: if, when an operation is performed, the current load factor is 1 1/x for some x, then the expected cost of the operation is O(x). Thus we can achieve the data locality of traditional linear probing without any of the disadvantages of primary clustering. One corollary is that, in the external-memory model with a data blocks of size B, graveyard hashing offers the following remarkably strong guarantee: at any load factor 1 1/x satisfying x = o(B), graveyard hashing achieves 1 + o(1) expected block transfers per operation. In contrast, past external-memory hash tables have only been able to offer a 1 + o(1) guarantee when the block size B is at least O(x2). Our results come with actionable lessons for both theoreticians and practitioners, in particular, that well- designed use of tombstones can completely change the asymptotic landscape of how the linear probing behaves (and even in workloads without deletions).  more » « less
Award ID(s):
1938180 2106999 2118620
NSF-PAR ID:
10298570
Author(s) / Creator(s):
; ;
Date Published:
Journal Name:
Annual Symposium on Foundations of Computer Science
ISSN:
0272-5428
Format(s):
Medium: X
Sponsoring Org:
National Science Foundation
More Like this
  1. For nearly six decades, the central open question in the study of hash tables has been to determine the optimal achievable tradeoff curve between time and space. State-of-the-art hash tables offer the following guarantee: If keys/values are Θ(logn) bits each, then it is possible to achieve constant-time insertions/deletions/queries while wasting only O(loglogn) bits of space per key when compared to the information-theoretic optimum. Even prior to this bound being achieved, the target of O(log log n) wasted bits per key was known to be a natural end goal, and was proven to be optimal for a number of closely related problems (e.g., stable hashing, dynamic retrieval, and dynamically-resized filters). This paper shows that O(log log n) wasted bits per key is not the end of the line for hashing. In fact, for any k ∈ [log∗ n], it is possible to achieve O(k)-time insertions/deletions, O(1)-time queries, and O(log(k) n) = Ologlog···logn 􏰟 􏰞􏰝 􏰠 k wasted bits per key (all with high probability in n). This means that, each time we increase inser- tion/deletion time by an additive constant, we reduce the wasted bits per key exponentially. We further show that this tradeoff curve is the best achievable by any of a large class of hash tables, including any hash table designed using the current framework for making constant-time hash tables succinct. Our results hold not just for fixed-capacity hash tables, but also for hash tables that are dynamically resized (this is a fundamental departure from what is possible for filters); and for hash tables that store very large keys/values, each of which can be up to no(1) bits (this breaks with the conventional wisdom that larger keys/values should lead to more wasted bits per key). For very small keys/values, we are able to tighten our bounds to o(1) wasted bits per key, even when k = O(1). Building on this, we obtain a constant-time dynamic filter that uses n􏰕logε−1􏰖+nloge+o(n) bits of space for a wide choice of 
    more » « less
  2. Hash tables are a ubiquitous class of dictionary data structures. However, standard hash table implementations do not translate well into the external memory model, because they do not incorporate locality for insertions. Iacono and Pătraşu established an update/query tradeoff curve for external-hash tables: a hash table that performs insertions in O(λ/B) amortized IOs requires Ω(logλ N) expected IOs for queries, where N is the number of items that can be stored in the data structure, B is the size of a memory transfer, M is the size of memory, and λ is a tuning parameter. They provide a complicated hashing data structure, which we call the IP hash table, that meets this curve for λ that is Ω(loglogM +logM N). In this paper, we present a simpler external-memory hash table, the Bundle of Arrays Hash Table (BOA), that is optimal for a narrower range of λ. The simplicity of BOAs allows them to be readily modified to achieve the following results: A new external-memory data structure, the Bundle of Trees Hash Table (BOT), that matches the performance of the IP hash table, while retaining some of the simplicity of the BOAs. The Cache-Oblivious Bundle of Trees Hash Table (COBOT), the first cache-oblivious hash table. This data structure matches the optimality of BOTs and IP hash tables over the same range of λ. 
    more » « less
  3. Despite being one of the oldest data structures in computer science, hash tables continue to be the focus of a great deal of both theoretical and empirical research. A central reason for this is that many of the fundamental properties that one desires from a hash table are difficult to achieve simultaneously; thus many variants offering different trade-offs have been proposed.

    This article introduces Iceberg hashing, a hash table that simultaneously offers the strongest known guarantees on a large number of core properties. Iceberg hashing supports constant-time operations while improving on the state of the art for space efficiency, cache efficiency, and low failure probability. Iceberg hashing is also the first hash table to support a load factor of up to1 - o(1)while being stable, meaning that the position where an element is stored only ever changes when resizes occur. In fact, in the setting where keys are Θ (logn) bits, the space guarantees that Iceberg hashing offers, namely that it uses at most\(\log \binom{|U|}{n} + O(n \log \ \text{log} n)\)bits to storenitems from a universeU, matches a lower bound by Demaine et al. that applies to any stable hash table.

    Iceberg hashing introduces new general-purpose techniques for some of the most basic aspects of hash-table design. Notably, our indirection-free technique for dynamic resizing, which we call waterfall addressing, and our techniques for achieving stability and very-high probability guarantees, can be applied to any hash table that makes use of the front-yard/backyard paradigm for hash table design.

     
    more » « less
  4. null (Ed.)
    RDMA has been an important building block for many high-performance distributed key-value stores in recent prior work. To sustain millions of operations per second per node, many of these works use hashing schemes, such as cuckoo hashing, that guarantee that an existing key can be found in a small, constant number of RDMA operations. In this paper, we explore whether linear probing is a compelling design alternative to cuckoo-based hashing schemes. Instead of reading a fixed number of bytes per RDMA request, this paper introduces a mathematical model that picks the optimal read size by balancing the cost of performing an RDMA operation with the probability of encountering a probe sequence of a certain length. The model can further leverage optimization hints about the location of clusters of keys, which commonly occur in skewed key distributions. We extensively evaluate the performance of linear probing with a variable read size in a modern cluster to understand the trade-offs between cuckoo hashing and linear probing. We find that cuckoo hashing outperforms linear probing only in very highly loaded hash tables (load factors at 90% or higher) that would be prohibitively expensive to maintain in practice. Overall, linear probing with variable-sized reads using our model has up to 2.8× higher throughput than cuckoo hashing, with throughput as high as 50M lookup operations per second per node. 
    more » « less
  5. Sparse linear algebra is an important kernel in many different applications. Among various sparse general matrix-matrix multiplication (SpGEMM) algorithms, Gustavson’s column-wise SpGEMM has good locality when reading input matrix and can be easily parallelized by distributing the computation of different columns of an output matrix to different processors. However, the sparse accumulation (SPA) step in column-wise SpGEMM, which merges partial sums from each of the multiplications by the row indices, is still a performance bottleneck. The state-of-the-art software implementation uses a hash table for partial sum search in the SPA, which makes SPA the largest contributor to the execution time of SpGEMM. There are three reasons that cause the SPA to become the bottleneck: (1) hash probing requires data-dependent branches that are difficult for a branch predictor to predict correctly; (2) the accumulation of partial sum is dependent on the results of the hash probing, which makes it difficult to hide the hash probing latency; and (3) hash collision requires time-consuming linear search and optimizations to reduce these collisions require an accurate estimation of the number of non-zeros in each column of the output matrix. This work proposes ASA architecture to accelerate the SPA. ASA overcomes the challenges of SPA by (1) executing the partial sum search and accumulate with a single instruction through ISA extension to eliminate data-dependent branches in hash probing, (2) using a dedicated on-chip cache to perform the search and accumulation in a pipelined fashion, (3) relying on the parallel search capability of a set-associative cache to reduce search latency, and (4) delaying the merging of overflowed entries. As a result, ASA achieves an average of 2.25× and 5.05× speedup as compared to the state-of-the-art software implementation of a Markov clustering application and its SpGEMM kernel, respectively. As compared to a state-of-the-art hashing accelerator design, ASA achieves an average of 1.95× speedup in the SpGEMM kernel. 
    more » « less