Distributed Computing: 21st International Symposium, DISC by Burkhard Monien, Karsten Tiemann (auth.), Andrzej Pelc

By Burkhard Monien, Karsten Tiemann (auth.), Andrzej Pelc (eds.)

Show description

Read Online or Download Distributed Computing: 21st International Symposium, DISC 2007, Lemesos, Cyprus, September 24-26, 2007. Proceedings PDF

Similar organization and data processing books

Languages and Compilers for Parallel Computing: 10th International Workshop, LCPC'97 Minneapolis, Minnesota, USA, August 7–9, 1997 Proceedings

This booklet constitutes the completely refereed post-workshop lawsuits of the tenth foreign Workshop on Languages and Compilers for Parallel Computing, LCPC'97, held in Minneapolis, Minnesota, united states in August 1997The publication provides 28 revised complete papers including 4 posters; all papers have been rigorously chosen for presentation on the workshop and went via a radical reviewing and revision section afterwards.

Cloud Computing: Web-basierte dynamische IT-Services (Informatik im Fokus) (German Edition)

Als Internetdienst erlaubt Cloud Computing die Bereitstellung und Nutzung von IT-Infrastruktur, Plattformen und Anwendungen. Dabei wird stets die aktuell benötigte Menge an Ressourcen zur Verfügung gestellt und abgerechnet. In dem Buch vermitteln die Autoren einen Überblick über Cloud-Computing-Architektur, ihre Anwendungen und Entwicklung.

Data Management in a Connected World: Essays Dedicated to Hartmut Wedekind on the Occasion of His 70th Birthday

Information administration platforms play the main an important function in development huge program s- tems. considering that glossy purposes are not any longer unmarried monolithic software program blocks yet hugely versatile and configurable collections of cooperative prone, the knowledge mana- ment layer additionally has to evolve to those new requisites.

Extra info for Distributed Computing: 21st International Symposium, DISC 2007, Lemesos, Cyprus, September 24-26, 2007. Proceedings

Sample text

That would reduce the overall storage overhead to O(log n) in Theorem 1, but for simplicity we will just assume that copies of d are stored. Notice that as long as the set of DoS-attacked nodes is static, our storage strategy could be transformed into an efficient insert protocol together with techniques in [2] to avoid congestion problems. However, for a dynamically changing set of DoS-attacked nodes this is tricky since some non-DoS-attacked nodes are now missing information that is necessary for our lookup protocol to work correctly.

Theorem 1 shows that the process converges before O(n log n) interactions with high probability. Suppose the process converges at some time τ = O(n log n). Then √ | = O( n log n) throughout this interval by Azuma’s inequality we have that |uτ − u0√ with high probability. So if u0 = u0 = ω( n log n), it follows that u0 ≥ u0 ≥ 0 throughout the execution, and in particular that the process does not terminate before convergence and that u is non-negative at convergence. But this excludes the y = n case, so the process converges to the initial majority value.

So we must also accept a small probability that we reach the blank configuration quickly, and the assurance that we reach it with probability 1 after a very long time. However, we can show that with high probability neither outcome occurs within a polynomial number of steps. Our technique is to adjust the potential functions used by the non-Byzantine process to account for Byzantine transitions. We then use these adjusted potential functions to show that (a) strong pressure exists to keep the process out of the high-b corner and in the high-x and high-y corners, and (b) the number of interactions (including Byzantine interactions) to reach the x or y corner is still small.

Download PDF sample

Rated 4.58 of 5 – based on 50 votes