Review - The Impact of DHT Routing Geometry on Resilience and Proximity

This paper offers an analysis of DHT geometries and shows that algorithmic
flexibility is desirable. Here, the focus is on studying static resilience,
the ability for a geometry to route when many nodes are unavailable, and
path latency. The authors find that having next-hop routing flexibility will
increase resilience, while having neighbor selection flexibility can
decrease latency (if the right neighbors are selected).

The analysis of several DHT geometries and algorithms is supplemented with a
categorization of methods presented in the literature. These include tree,
hypercube, ring, etc. and serve as a base for the discussion. Using
simulation Gummandi et al. conclude that that the selection of routing
geometry is essential, because it constrains other design issues. Further,
the most important difference between geometries is the degree of
flexibility they offer. Their analysis shows that the ring design provides
the most flexibility, which allows greater resilience to failure and can
potentially lead to the selection of low latency neighbors.

While this paper was interesting, is it unclear how much resilience is
needed in practical cases. I can believe that by reducing latency, we can
achieve better performance. However, static resilience does not seem to be
an issue if the algorithm implements an efficient recovery. DHTs inherently
provide good recovery. The authors should have outlined a need for static
resilience (it's not likely that half the Internet will go down at any given
time).

As a final note, the graphs in this paper as ridiculous. In many cases it is
impossible to distinguish one line from another. This made analysis of the
provides figures impossible.
Received on Mon Nov 21 2005 - 02:18:10 EST