| | |
Summary: Resettable Vector Clocks: A Case Study in Designing
Graybox Fault-Tolerance
By
Murat Demirbas, M.S.
The Ohio State University, 2000
Professor Anish Arora, Adviser
The task of designing fault-tolerance for large-scale applications (applications that
inevitably contain multiple components) can be signi cantly simpli ed by designing
fault-tolerance at the component level. In contrast to the traditional whitebox and
blackbox methods, a graybox method for designing fault-tolerance to components al-
lows the design of scalable and low-cost fault-tolerance by exploiting the contracts of
the components.
In this thesis, we present a case study in designing graybox fault-tolerance. Our
case study focuses on designing bounded-space fault-tolerance for applications that
use vector clocks (VC). To this end, we generalize the notion of VC to resettable
vector clocks (RVC) and identify a contract under which a bounded-space RVC im-
plementation can be substituted for VC in client applications without a ecting the
client's correctness. Further, we design fault-tolerance to bounded-space RVC by
exploiting its contract. Based on this bounded-space and fault-tolerant RVC, we
1
|
