Article:		Distributed Systems Topologies: Part 2
Subject:		The problem of decidability
Date:		2002-01-10 06:23:46
From:		hhh
I think the ability to decide whether an event has happend or not, is very important because many application rely on this ability (ex. Databases, DNS, DSM, etc.). Decentralized systems often lag this ability, because the only way to decide anything is to search througth all nodes. Imagine searching through all of Gnutella to decide whether a file exits - this is next to imposible (I didn't say imposible :o). This property is part of the 'Information coherence' property, but I think it should be a specific part of the analysis. Currently I am doing my master's thesis, and I have had good results using a hierarchical topology with a decentralized topology as a safety net to increase fault-tolerance. This is all in its early stages, but the hierarchical topology gives decidability and preliminary tests show good scalability. Hans Henrik Happe

Showing messages 1 through 2 of 2.

• The problem of decidability
  2002-01-16 03:30:36  kiwipeso [Reply | View]
  
  
  I'm making a protocol (Samizdat) that is Decentralized Hierarchy based, it uses a list of files & metadata fors peers that is sent to servers on the network.
  Only servers need to check the user lists to find files, it can also send you the closest or fastest connection to the file you want.
  This is a key part of the operating system I am developing (KAOS)

• The problem of decidability
  2002-01-10 10:17:02  Nelson Minar [Reply | View]
  
  
  Thanks for the thoughts! I had a hard time defining "information coherence" for this article. I agree, "decidability" is a well-defined and well studied concept that gets at some of the issues of coherence.
  
  Academic research on distributed systems has done a huge amount of work on the issues my article is about. My article doesn't really do that work justice; for folks who want to know more, there's a lot to learn!