The paper is highlighting a significant deviation in peer-to-peer systesm
from the right direction : The main problem of real scalability for robust
internet-scale storage is not to procide guarantees for lookups, but rather
is to provide guarantees for redundany maintenance in the current setting of
Internet.
The paper uses a conservative model for peer-to-peer systems' host arrival
and departure to infer minimum participation requirement for having a
certain level of redundancy. The result show that only with an extremely
large cross-system bandwidth, one can have high redundacy, data and
dynamics. Although the work does not suggest applicable solutions to the
problems it is capturing, it has a significant contribution in the sense
that it guides the future focus of research in the field.
The authors could also argue on the success of the current p2p systems like
Gnutella despite their argument on the scarce availability of data
--considering the high dynamics of such networks. In fact, what makes the
widely deployed p2p systems socially succesfful may not be in the strength
of the system to provide exact redundancy, but rather lower levels of
redundacy (like : the system has always enough songs from a certain artist)
might be enough to make it attractive.
Received on Mon Nov 14 2005 - 11:12:12 EST
This archive was generated by hypermail 2.2.0 : Mon Nov 14 2005 - 11:12:13 EST