|
The capability of network configuration solution to record its past experience, and use it to infer
the future updates required for its software at runtime, in response to internal or
external triggers, imply that the NC is capable of evolution. Accordingly a network
built of NCs will also be capable of evolution as the network behavior is perceived as
the collective behavior of its constituents. We envision the NC experience to be
recorded in terms of perceived behavioral patterns, responses to a perceived pattern or
a sequence of patterns represented as sequence of updates, and the resulting outcome
of each of the adopted decisions in terms of the overall gain in cell performance. The
inference engine in the NC's RU will be responsible to pick up the most suitable
course of action based on the recorded experience and present NC states. Since each
cell can undergo changes independent of its peer cells we need a mechanism to
guarantee that evolving cells will maintain communication in spite of any changes
applied to their software To allow the independent evolution of cells while
maintaining continuous operation of the network, a dynamic interface management
technique is required to insure that progressive evolution of cells does not lead to
network breakage. A cell publishes one or more local interfaces exposing its services
to other cells in the network. When a cell evolution induces changes in the cell local
interface, the cell publishes a new interface to support its evolved behavior while
retaining the old interfaces to allow other cells that didn't undergo a compatible
evolution to continue using the cell services through the older interfaces.
|
