Cuckoo Trie: Exploiting Memory-Level Parallelism for Efficient DRAM Indexing

Adar Zeitak; Adam Morrison

doi:10.1145/3477132.3483551

Cuckoo Trie: Exploiting Memory-Level Parallelism for Efficient DRAM Indexing

Adar Zeitak, Adam Morrison

School of Computer Science

Tel Aviv University

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

10 Scopus citations

Abstract

We present the Cuckoo Trie, a fast, memory-efficient ordered index structure. The Cuckoo Trie is designed to have memory-level parallelism - -which a modern out-of-order processor can exploit to execute DRAM accesses in parallel - - without sacrificing memory efficiency. The Cuckoo Trie thus breaks a fundamental performance barrier faced by current indexes, whose bottleneck is a series of dependent pointer-chasing DRAM accesses - -e.g., traversing a search tree path - - which the processor cannot parallelize. Our evaluation shows that the Cuckoo Trie outperforms state-of-the-art-indexes by up to 20%-360% on a variety of datasets and workloads, typically with a smaller or comparable memory footprint.

Original language	English
Title of host publication	SOSP 2021 - Proceedings of the 28th ACM Symposium on Operating Systems Principles
Publisher	Association for Computing Machinery, Inc
Pages	147-162
Number of pages	16
ISBN (Electronic)	9781450387095
DOIs	https://doi.org/10.1145/3477132.3483551
State	Published - 26 Oct 2021
Event	28th ACM Symposium on Operating Systems Principles, SOSP 2021 - Virtual, Online, Germany Duration: 26 Oct 2021 → 29 Oct 2021

Publication series

Name	SOSP 2021 - Proceedings of the 28th ACM Symposium on Operating Systems Principles

Conference

Conference	28th ACM Symposium on Operating Systems Principles, SOSP 2021
Country/Territory	Germany
City	Virtual, Online
Period	26/10/21 → 29/10/21

Funding

Funders	Funder number
Blavatnik Family Foundation
Israel Science Foundation	2005/17

Keywords

MLP
index
memory-level parallelism
trie

Access to Document

10.1145/3477132.3483551

Cite this

Zeitak, A., & Morrison, A. (2021). Cuckoo Trie: Exploiting Memory-Level Parallelism for Efficient DRAM Indexing. In SOSP 2021 - Proceedings of the 28th ACM Symposium on Operating Systems Principles (pp. 147-162). (SOSP 2021 - Proceedings of the 28th ACM Symposium on Operating Systems Principles). Association for Computing Machinery, Inc. https://doi.org/10.1145/3477132.3483551

@inproceedings{3cd0f3dcdbb340d2a99184ef7ad10352,

title = "Cuckoo Trie: Exploiting Memory-Level Parallelism for Efficient DRAM Indexing",

abstract = "We present the Cuckoo Trie, a fast, memory-efficient ordered index structure. The Cuckoo Trie is designed to have memory-level parallelism - -which a modern out-of-order processor can exploit to execute DRAM accesses in parallel - - without sacrificing memory efficiency. The Cuckoo Trie thus breaks a fundamental performance barrier faced by current indexes, whose bottleneck is a series of dependent pointer-chasing DRAM accesses - -e.g., traversing a search tree path - - which the processor cannot parallelize. Our evaluation shows that the Cuckoo Trie outperforms state-of-the-art-indexes by up to 20%-360% on a variety of datasets and workloads, typically with a smaller or comparable memory footprint.",

keywords = "MLP, index, memory-level parallelism, trie",

author = "Adar Zeitak and Adam Morrison",

note = "Publisher Copyright: {\textcopyright} 2021 ACM.; 28th ACM Symposium on Operating Systems Principles, SOSP 2021 ; Conference date: 26-10-2021 Through 29-10-2021",

year = "2021",

month = oct,

day = "26",

doi = "10.1145/3477132.3483551",

language = "אנגלית",

series = "SOSP 2021 - Proceedings of the 28th ACM Symposium on Operating Systems Principles",

publisher = "Association for Computing Machinery, Inc",

pages = "147--162",

booktitle = "SOSP 2021 - Proceedings of the 28th ACM Symposium on Operating Systems Principles",

}

Zeitak, A & Morrison, A 2021, Cuckoo Trie: Exploiting Memory-Level Parallelism for Efficient DRAM Indexing. in SOSP 2021 - Proceedings of the 28th ACM Symposium on Operating Systems Principles. SOSP 2021 - Proceedings of the 28th ACM Symposium on Operating Systems Principles, Association for Computing Machinery, Inc, pp. 147-162, 28th ACM Symposium on Operating Systems Principles, SOSP 2021, Virtual, Online, Germany, 26/10/21. https://doi.org/10.1145/3477132.3483551

Cuckoo Trie: Exploiting Memory-Level Parallelism for Efficient DRAM Indexing. / Zeitak, Adar; Morrison, Adam.
SOSP 2021 - Proceedings of the 28th ACM Symposium on Operating Systems Principles. Association for Computing Machinery, Inc, 2021. p. 147-162 (SOSP 2021 - Proceedings of the 28th ACM Symposium on Operating Systems Principles).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - Cuckoo Trie

T2 - 28th ACM Symposium on Operating Systems Principles, SOSP 2021

AU - Zeitak, Adar

AU - Morrison, Adam

PY - 2021/10/26

Y1 - 2021/10/26

N2 - We present the Cuckoo Trie, a fast, memory-efficient ordered index structure. The Cuckoo Trie is designed to have memory-level parallelism - -which a modern out-of-order processor can exploit to execute DRAM accesses in parallel - - without sacrificing memory efficiency. The Cuckoo Trie thus breaks a fundamental performance barrier faced by current indexes, whose bottleneck is a series of dependent pointer-chasing DRAM accesses - -e.g., traversing a search tree path - - which the processor cannot parallelize. Our evaluation shows that the Cuckoo Trie outperforms state-of-the-art-indexes by up to 20%-360% on a variety of datasets and workloads, typically with a smaller or comparable memory footprint.

AB - We present the Cuckoo Trie, a fast, memory-efficient ordered index structure. The Cuckoo Trie is designed to have memory-level parallelism - -which a modern out-of-order processor can exploit to execute DRAM accesses in parallel - - without sacrificing memory efficiency. The Cuckoo Trie thus breaks a fundamental performance barrier faced by current indexes, whose bottleneck is a series of dependent pointer-chasing DRAM accesses - -e.g., traversing a search tree path - - which the processor cannot parallelize. Our evaluation shows that the Cuckoo Trie outperforms state-of-the-art-indexes by up to 20%-360% on a variety of datasets and workloads, typically with a smaller or comparable memory footprint.

KW - MLP

KW - index

KW - memory-level parallelism

KW - trie

UR - http://www.scopus.com/inward/record.url?scp=85119098408&partnerID=8YFLogxK

U2 - 10.1145/3477132.3483551

DO - 10.1145/3477132.3483551

M3 - ???researchoutput.researchoutputtypes.contributiontobookanthology.conference???

AN - SCOPUS:85119098408

T3 - SOSP 2021 - Proceedings of the 28th ACM Symposium on Operating Systems Principles

SP - 147

EP - 162

BT - SOSP 2021 - Proceedings of the 28th ACM Symposium on Operating Systems Principles

PB - Association for Computing Machinery, Inc

Y2 - 26 October 2021 through 29 October 2021

ER -

Cuckoo Trie: Exploiting Memory-Level Parallelism for Efficient DRAM Indexing

Abstract

Publication series

Conference

Funding

Keywords

Access to Document

Other files and links

Fingerprint

Cite this