Institutt for informatikk 
Univ. of Tromsø

Faculty of Science
 
Information
The Department
Studies
Bachelor
Master (2 years)
Master (5 years)
Courses Spring 2007
Courses B and M
Courses PhD
CS-lunsj
Eksamensoppgaver
International students
Educational quality
CS library
Research
Employees
 
Services
Weather observations
Technical services
Adm. services
Departments
Biology
Physics
Geology
Chemistry
Tromsø Geophysical Observatory
Mathematics and Statistics
 
Language:

 
[ Login ]
set width of left cell in netscape 4
Startpage › matnat › informatikk › studier › gjester › 45
Printer friendly

Issues with Protocols and Dependent Failures


26.03.2004

Foredragsholdere
Keith Marzullo

Tid: Fredag 10:15 til 12:00
Sted: Lille auditorium Realfagsbygget

Title: The mechanics of paper submission and review.

Speaker: Keith Marzullo, UC San Diego, USA and Department of Computer Science, UiTø.

Abstract:
Often, lower bounds are specified using what has been called a threshold model: out of n components, no more than t of them are assumed to be faulty. Protocols that are optimal with respect to such a model, however, tacitly assume that failures are independent and identically distributed (IID) since any set of t or fewer components can be faulty.

We have developed a model for representing non-IID failures of processes and explored the design of protocols using our model. I will present our model and give some examples of how it can be used to represent non-IID failures. I will then give some bounds, expressed in our model, that state when there is sufficient process replication to support a protocol that requires in the threshold model n > kt for some integer k > 1. I will also show what replication bounds mean in our model where, in the threshold model, the bound is expressed as n > at/b for integer a > b > 0 and illustrate them with a novel leader election protocol (first posed but not generally solved by Navin Budhiraja). I will finally consider the following problem: what if one has a protocol P that requires n > kt, and a set of processors that has enough replication to solve the same problem that P solves, but not sufficient replication to run protocol P? I show how to transform P into a protocol that can be run on the existing system.

The work in this talk is joint work with Flavio Junqueira.


Eventuelle spørsmål i tilknytting til Marzullos forelesning kan rettes til professor Dag Johansen, tlf 7764 4047.

Vel møtt!


Institutt for informatikk, det matematisk- naturvitenskapelige fakultet,
Universitetet i Tromsø, 9037 Tromsø
Updated by konsulent Jan Fuglesteg on 25.03.2004 at 14:20
Ansvarlig redaktør: adm.leder Svein Tore Jensen


Read this page in: Bokmål

set width of main cell in netscape 4

All UiT-pages
Advanced search

Pages:
Showing 1-15 of 67
   
The Ghost of GUIDE
Middelware for QoS...
Life Sciences Research on Blue Gene/L
Challenges in non-blocking synchronization
Introduction to the modelling and simulation of clinical processes with the Arena tool
Model Driven Architecture in HealthCare
Communication theory down to everyday medicine
Sensor and transducers on the human body
Open Code - a revolution in the future computer systems market
Disease surveillance system
Building a Requirements Specification - Starting a New Telemedicine Project
Reactive Universal Constructions for Non-blocking Synchronization
Privacy and Background Knowledge
Multimedia Search & Retrieval, Finding
Hvad har opera til fælles med patient journal og hvad har det med large displaywall at gøre?
set width of right cell in netscape 4