This shows you the differences between two versions of the page.

Link to this comparison view

decentralized_storage_systems:soaadditional [2012/04/23 01:10] (current)
julia created
Line 1: Line 1:
 +====== Additional Functionality to the "State of Art" System used by VC ======
 +Start up paper to discover the one of the possibilities is [[http://www.cs.rice.edu/Conferences/IPTPS02/170.pdf|Erasure coding vs. replication: A quantitative comparison]].
 +Erasure Coding - provide redundancy without the overhead of pure replication. The main idea behind this method: divide an object into //m// parts and decode them into //n// parts, where //n// > //m//.  Reasoning to use Erasure Coding: provide way much less bandwidth and storage consumption for providing similar level of durability as replicated system.
 +Overall, erasure-resilience is better to utilized for durability, when replication could be better utilized for latency reduction. Hence, replication should be applied smartly to the smaller files and erasure coding to the bigger files.
 +Currently still the question is how to reduce overhead from data fragment recovery. New solution is to apply network coding which could be found in the [[http://ieeexplore.ieee.org/xpl/articleDetails.jsp?tp=&arnumber=4215814&contentType=Conference+Publications&queryText%3DNetwork+coding+for+distributed+storage+systems|Network Coding for DSS]]. This solution should be the best for the VC application as churn will be presented. 
 +What could be done even more to provide more availability in uncertain conditions is to classify data types and according to it choose different fault-tolerance techniques. One fo the options could be [[http://www.kom.tu-darmstadt.de/~abboud/APK+11-1.pdf|P2P video streaming]] technique application for the video files. What does it mean?  Several independent so-called descriptions will be created from a video file. This method is called Multiple Description Coding (MDC). Each description is a video segment that contributes a certain amount to the video quality and can be decoded independently. The more descriptions are received, the higher is the received quality. Hence, each data file could be divided into n amount of pieces, e.g. odd and even horizontal and vertical frames for the video file, and then erasure coding/network coding could be applied to each of the piece itself.
 +This approach will provide order of magnitude more durability, availability and fault-tolerance then replication.
decentralized_storage_systems/soaadditional.txt · Last modified: 2012/04/23 01:10 by julia
Except where otherwise noted, content on this wiki is licensed under the following license: CC Attribution-Share Alike 3.0 Unported
Recent changes RSS feed Donate Powered by PHP Valid XHTML 1.0 Valid CSS Driven by DokuWiki