LeraoSoft - Knowledge Distributed

Spent five hours doing reserch on other DHT implimentations developed by diffrent people. Had little luck getting most to install esally or run at all.

A few promising items came out of it that i am going to further investigate. The projects i’m looking at now are

• Chimera
• Pastry
• Bamboo

Each has it’s own advantages and disadvantages, for example CHimera is simple but low level, bamboo is a pain to set up and complex code however Pastry and the Scribe extensions seem to be promising.

Today i had an appifiny, after meeting with my advisor i discoverd the optimal way for the network to work and i beleve a solution that will let me impliment it.  That methood is DHT and the solution is Bamboo .

DHT’s are the appropriate solution becasue they will allow our project to maintain some structure withought needing to broadcast the entire file to all nodes.  Also DHT’s provide a network overlay, routing algorithums and a network protocall.  This simplifies my work from the privious approch wich was to use Nemo which would only provide a network overlay.

Bamboo is great because of it’s simple protocol of Get and Put as well as it’s efficiency in handling changing networks with high churn(or nodes leaving and joining often).  It is also Open-source under the GPL and  very modular for the changes we will make.

Current concepts are to have a hash key for each block and use these hash keys as channels for data transfer.  More information on this later.

I was looking at Northwestern for college and discovered AquaLab a research group working on distributed computing. Of particular interest to me is the Nemo project. I also believe i can make it work in reverse allowing for a great distribution system for file fragments. Possibly even using the cluster like system for distribution to do work in creating parity data.