Power of d choices with simple tabulation

TY - GEN

T1 - Power of d choices with simple tabulation

AU - Aamand, Anders

AU - Knudsen, Mathias Bæk Tejs

AU - Thorup, Mikkel

PY - 2018/7/1

Y1 - 2018/7/1

N2 - We consider the classic d-choice paradigm of Azar et al. [STOC'94] in which m balls are put into n bins sequentially as follows: For each ball we are given a choice of d bins chosen according to d hash functions and the ball is placed in the least loaded of these bins, breaking ties arbitrarily. The interest is in the number of balls in the fullest bin after all balls have been placed. In this paper we suppose that the d hash functions are simple tabulation hash functions which are easy to implement and can be evaluated in constant time. Generalising a result by Dahlgaard et al. [SODA'16] we show that for an arbitrary constant d ≥ 2 the expected maximum load is at mostlg lg lg d n + O(1). We further show that by using a simple tie-breaking algorithm introduced by Vöcking [J.ACM'03] the expected maximum load is reduced tod lg lg lg ϕ n d + O(1) where ϕd is the rate of growth of the d-ary Fibonacci numbers. Both of these expected bounds match those known from the fully random setting. The analysis by Dahlgaard et al. relies on a proof by P tra cu and Thorup [J.ACM'11] concerning the use of simple tabulation for cuckoo hashing. We require a generalisation to d > 2 hash functions, but the original proof is an 8-page tour de force of ad-hoc arguments that do not appear to generalise. Our main technical contribution is a shorter, simpler and more accessible proof of the result by P tra cu and Thorup, where the relevant parts generalise nicely to the analysis of d choices.

AB - We consider the classic d-choice paradigm of Azar et al. [STOC'94] in which m balls are put into n bins sequentially as follows: For each ball we are given a choice of d bins chosen according to d hash functions and the ball is placed in the least loaded of these bins, breaking ties arbitrarily. The interest is in the number of balls in the fullest bin after all balls have been placed. In this paper we suppose that the d hash functions are simple tabulation hash functions which are easy to implement and can be evaluated in constant time. Generalising a result by Dahlgaard et al. [SODA'16] we show that for an arbitrary constant d ≥ 2 the expected maximum load is at mostlg lg lg d n + O(1). We further show that by using a simple tie-breaking algorithm introduced by Vöcking [J.ACM'03] the expected maximum load is reduced tod lg lg lg ϕ n d + O(1) where ϕd is the rate of growth of the d-ary Fibonacci numbers. Both of these expected bounds match those known from the fully random setting. The analysis by Dahlgaard et al. relies on a proof by P tra cu and Thorup [J.ACM'11] concerning the use of simple tabulation for cuckoo hashing. We require a generalisation to d > 2 hash functions, but the original proof is an 8-page tour de force of ad-hoc arguments that do not appear to generalise. Our main technical contribution is a shorter, simpler and more accessible proof of the result by P tra cu and Thorup, where the relevant parts generalise nicely to the analysis of d choices.

KW - Balls

KW - Bins

KW - Load Balancing

KW - Phrases Hashing

KW - Simple Tabulation

UR - http://www.scopus.com/inward/record.url?scp=85049793862&partnerID=8YFLogxK

U2 - 10.4230/LIPIcs.ICALP.2018.5

DO - 10.4230/LIPIcs.ICALP.2018.5

M3 - Article in proceedings

AN - SCOPUS:85049793862

T3 - Leibniz International Proceedings in Informatics, LIPIcs

BT - 45th International Colloquium on Automata, Languages, and Programming, ICALP 2018

A2 - Kaklamanis, Christos

A2 - Marx, Daniel

A2 - Chatzigiannakis, Ioannis

A2 - Sannella, Donald

PB - Schloss Dagstuhl- Leibniz-Zentrum fur Informatik GmbH, Dagstuhl Publishing

T2 - 45th International Colloquium on Automata, Languages, and Programming, ICALP 2018

Y2 - 9 July 2018 through 13 July 2018

ER -