Application Framework for Programmable Network Control - Advances in Network-Embedded Management and Applications - page 41

adaptation process may
process but may also be
edge weights of a shortest path routing protocol
typically consists of (1) infer
(2) calculating network state (3) and
that leads towards the desired
known model in control theory to influence the behavior of a dynamic system
[35], provides a minimal framework for network control
In order to match the network to a state that is optimal to an application, the
application has to collect (
application developer chooses application specific abstractions
update a model the application uses internally
information from all or a subset of NEs to update the internal model. I
principle, the internal

information, such as computing or hosting
service level agreements.
The control application
needed to adjust the network be
application needs (e.g. a stable, optimized state), which are described by the
reference. To implement changes in the network, the control application
translates decisions into
specific to each NE involved in the application. This means that
needs to provide a distributed transaction monitor to keep network
manipulations that involve multiple NE

Fig. 2. The application framework to control

In control theory, a measurement
subtracted from a reference value
the control application. In our framework,
that represent network state
controlled state (AC Actions). For example, a controller may manipulate edge
weights in shortest path routing based on throughput information. Such a
scenario is meaningful if the relat
(Ō) is known or can be learnt
traffic to avoid congestion, for example