Abstract
Overlay networks emerged in the last decade and generate now a large portion of the
Internet traffic Internet07. Even though they were initially associated exclusively
with copyright infringements and, thus, provoked considerable public odium, their
good properties supersede and the overlay networking technology is here to stay. On
one hand, overlay networks can provide specific applications without being supported
by current underlay network, e.g., multicast. On the other hand, the client-server
model does not scale in different respects, and many content distributors find it more
efficient to deliver their content using overlay networks. For example, free Linux
distributions (e.g., Red Hat’s Fedora or Debian) are distributed via BitTorrent.
Furthermore, the Internet traffic stemming from overlay-based applications, e.g.,
Peer-to-Peer (P2P) applications, increases rapidly with the increase of available
bandwidth of end-nodes. Therefore, a slow paradigm shift from centrally offered
services to services offered by end-nodes is in progress. For today’s Telecommunication
Service Providers (telcos) and Internet Service Providers (ISP) the key issue
arising is how to control and manage network traffic stemming from such overlaybased
applications. As the structure of overlays determines traffic flows in ISP
networks, it is highly efficient for an ISP to influence the overlay configuration based
on information on their structure. Thus, the management of overlays aims (a) to
maximize the benefit for multiple operators/ISPs involved and (b) to increase the
capability to withstand faults and balancing the network load.
Traditional traffic engineering methodologies are insufficient to satisfy simultaneously
the provider’s goal to maximize network usage and the user’s goal to
maximize his utility, since most of current overlay networks are deployed oblivious of
the underlying physical network Awduche99. E.g., traffic between two nodes in the
Telefonica network in Madrid can go via Tokyo, incurring large and unnecessary
costs. This is why network operators show a strong interest in influencing overlay
network operations and shaping the overlay traffic as to comply with their needs.
Within that context, SmoothIT proposes a new traffic management mechanism
termed Economic Traffic Management (ETM), which provides for an incentivecompatibility
in those interactions foreseen between overlay applications and the
underlying ISP networks in order to gain the following measurable impacts:
1. Cost saving for ISPs: lower operation costs, due to ETM-based traffic engineering,
lower interconnection costs, since traffic can be kept inside an ISP’s domain, and
lower capacity extension cost, since capacity requirements can be forecasted with
much higher accuracy.
2. Lower prices for end-users, due to competitive pricing by the ISP, which are
enabled by new ETM mechanisms.
3. Better Quality-of-Service (QoS) for overlay-based applications across ISP
domains, due to the usage of ETM-based traffic engineering. This leads to an
improved media consumption experience for end users.
Users
Please
log in to take part in the discussion (add own reviews or comments).