Advanced Search

Browse by Discipline

Scientific Societies

E-print Alerts

Add E-prints

E-print Network

  Advanced Search  

Adaptive algorithms utilyzing adaptive collect and snapshot

Summary: Adaptive algorithms utilyzing adaptive collect
and snapshot
Yehuda Afek \Lambda Eli Gafni y Michael Merritt z
August 25, 1999
Several adaptive algorithms are automatically generated via a simple transformation from
single­writer multi­reader algorithms, using the O(k) adaptive collect algorithm of Attiya and
Fouren [AF98a]. Among these algorithms are an adaptive snapshot algorithm with step complex­
ity O(k 2 ), and three algorithms solving (2k \Gamma 1)­renaming, but with high step complexities (k4 k ,
\Omega\Gamma/ k=2 ), and k 3 ), where k is the contention, the number of processes actually taking steps during
the run of the algorithm. The transformation does not always produce an adaptive algorithm:
the O(n log n) lattice­agreement (one shot snapshot) algorithm of Attiya and Rachman [AR93]
is one example. However, we show that a simple modification of the original algorithm allows
the transformation to produce an adaptive, O(k log k) lattice­agreement algorithm, matching the
best­known step complexity of Attiya and Fouren's algorithm [AF98].
Finally, we present a specially­tailored algorithm for (2k \Gamma 1)­renaming that uses any of these
adaptive lattice­agreement or snapshot algorithms as a component. Building on a renaming
algorithm of Attiya and Fouren, the adaptive (2k \Gamma 1)­renaming algorithm has step complexity
O(k 2 ), which is better than any other known adaptive algorithm for optimal renaming.
1 Introduction


Source: Afek, Yehuda - School of Computer Science, Tel Aviv University


Collections: Computer Technologies and Information Sciences