Sofi-servicearchitectures-081210 by xiaoyounan


									             Integrating SSA&I projects into the Future Internet activities

                                   A fundamental limitation of the
                                           current Internet for the
                                       future Internet of Services:
    Addressing and routing services
                                           Input to the FIA Ghent,
                                               16 December 2010

                                       From SOFI invited expert:
                                                 Luciano Baresi
             Abstracted Service DNS
• Fundamental limitation: current service-to-service addressing is strongly
     based on physical location (is part of the uri!) and technology (how shall
     I address a service in a mobile phone?)
• Current Internet: interactions between distributed components are
     limited by technologies that are static (in structure and even on physical
     set-up), difficult to change and prone to failure. Development is strongly
     influenced by the technology platform (web, mobile net…)
• How it should be: development (and most of deployment) of distributed
     solutions should be made in a way that is independent to the physical
     location and underlying technology used. e.g. In the cloud, we should be
     able to move services from one location to another and the existing
     compositions should keep working.
• What is needed: a service addressing mechanism that is independent
10.04.2013 the physical location (and if possible, technology) of the services.

                Scalable Service DNS
• Option 1: to have a Service ID that identifies it worldwide, independent
  from the location. Then the underlying levels are the ones responsible of
  routing the messages to the current location of the services as it is done in
  mobile phones: we don‘t need to know where the phone is, only use the
  full phone number with its international code to reach it).
• A more extreme option would be to do service addressing not by a world-
  wide unique id, but directly by some sort of service description. But that
  may not be scalable.
• Detachment of service addressing from the URIs is specially important if
  the other computational entities in the FI (things, objects, sensors,
  resources) are going to be described as services.
• This last point also brings a very important aspect: new service
  descriptions that can describe all the computational entities in the FI.

               Content-based Routing
•   Routing based on destination addresses can also lead to issues and problems, such
    as attacks, difficulties in handling mobilility, band overload and misuse (data must
    arrive before figuring out if it is useful or not).
•   With content-based routing (or even better if based on the interest on the
    content), i.e. by exploiting publish and subscribe mechanisms, some of those
    issues could be solved or reduced.
•   Regarding higher layers, the Network of the Future should replace (or
    complement) the current data-oriented management with solutions more
    oriented to services and things, even at the layers below application. In other
    words, things and services should be concepts modelled and managed at the
    layers between IP and application (included).
•   Are things and services just URIs? Could be useful to consider them as something
    else or more than a URI? It would be useful to provide the application level with
    more complex or meaningful abstractions. This could also benefit the routing and
    other network operations based on those lower levels.


To top