Review: An Architecture for Content Routing Support in the Internet

The paper identifies that the primary use of the Internet is content
delivery. The authors claim that current content routing schemes will
not scale to handle the increasing global demands. They describe a
name-based content routing design as part of an explicit content layer
in the Internet.
The authors begin by outlining the problems with current content routing
schemes, including high latencies as a result of several RTTs to locate
the content server and inaccessibility to the information needed for
content routing. They also say that proprietary approaches violate the
basic philosophy of using open, standard protocols.
The paper then describes the proposed network integrated content
routing, which consist of:
1. Content Routers (CR) to provide support in the Internet core to
distribute, maintain and make ise of information about content delivyer.
They allow fast location of nearby content replicas using naming
information. CRs act as noth conentional IP routers too. Firewalls,
gateways and BGP routers can be used for this purpose.
2. Name-Based Routing Protocol (NBRP) to distribute the names and set
the routing tables to provide routing in a structure similar to BGP.
3. Internet Resolution Protocol (INRP) to perform efficient lookup on
the distributed integrated name-based routing system.
The authors implemented the above protocol in C++ and provided some
experimental results, which showed that the client name lookup is faster
and the latency is less variable. The memory required, although high, is
acceptable.
The paper briefly describes how to deploy their scheme. It only requires
implementing content routers on top of existing routing hardware, and
implementing INRP and NBRP in ISP name servers.
The paper brings forth a valid point, however, the authors have not
provided convincing arguments to support the scalability of their
name-based routing scheme. They discuss name-based aggregation, but it
is unclear how well this works compared to IP aggregation, despite their
attempt at providing results to show its efficiency. On the whole, the
idea presented is interesting.
Received on Wed Nov 29 2006 - 17:20:02 EST