[CSC2231] Paper Review: High Availability, Scalable Storage, Dynamic Peer Networks: Pick Two

This paper studies the availability of peer-to-peer storage systems and
comes up with a mathematical model to determine the effect of various
parameters on availability.

I am not a fan of mathematical model so I find the discussion section
more interesting, especially for the admission control. It seems very
intuitive to delay a recently-joined client from sharing the load
because the system does not know about the availability of this client.
The system should slowly shift the load to such a client only if the
client has joined long enough. This technique can also prevent the loss
of data in an attack when enough clients join the system, obtain their
share of load, and then leave the system.

It seems that this paper suggests a system that allows me to replicate
my entire harddrive to a P2P system. I feel a strong disincentive to
make all data distributed at anonymous hosts that may or may not have
interest in my data. First, there is privacy on my personal data.
Second, even if the data is not private, I find unethical to use others'
machines to store my stuffs unless they are also interested in it (such
as multimedia files). So, personally, I think this type of P2P systems
only make sense in individual organizations that wants to make use of
all resources of their computers.
Received on Mon Nov 14 2005 - 10:34:44 EST