Review - RON

Review of Resilient Overlay Network
by Ivan Hernández

The paper presents Resilient Overlay Network (RON). RON will enable a
group of computers to communicate with each other in the face of
problems with the underlying Internet paths connecting them. The RON
is an alternative to build distributed applications that have specific
quality of service requirement. The authors say that the current
Internet may not meet this requirements, mainly because the bad BGP's
convergence time.

RON is formed from a set of N computers that will join the overlay
network. This N computers have the knowledge of the existence of each
other, they use virtual links to form a mesh of N nodes. RON takes
advantage from the fact that between any two computers in the overlay
network is highly probable that there is physical path
redundancy. Thanks to this, RON is able to provide fast convergence
and continuous communication between any to computers in RON, if there
are problems with the underlying Internet paths connecting them.

Different applications may have different service requirements to
operate. RON allow applications to influence the choice of paths using
a custom metric, for instance, latency, throughput, and packet loss
rate. A node in RON exchanges its metrics information -- per type of
application -- to the other nodes using routing messages.
Additionally, RON aims to provide fine-grained policy routing, this is
policies aimed at user or hosts; the RON routers will build a separate
routing table for each of these policies. When a packet arrives to a
RON router, the router will identify the packet policy and will select
an adequate routing table, then the router will use the adequate
metric for the packet and will determine the next-hop in the overlay
network.

The authors implement an IP forwarded to send IP traffic over the RON
and emit it at the other end. The evaluation of the prototype show
improvements in the convergence time with respect to BGP, furthermore,
in one of the datasets the prototype was able to route around all the
outage situations. The prototype was very useful in an experiment that
simulated a flood-induced outage, the prototype was able to route
around the problem. The paper is fine. One interesting point in RON is
the performance database. The flaw in the solution is that it can not
be used on a large-scale application, because it relies on a mesh
architecture.
Received on Sat Oct 14 2006 - 18:07:59 EDT