Facilitating Adaptive Placement of Management and Control Functions in Converged ICT Systems - Advances in Network-Embedded Management and Applications - page 62

be also recompiled and redeployed to programmable network interfaces,
which is beginning to be applied in virtualized systems (e.g. [32]).
A major feature of the M&C framework is the support for deployment on
different levels. While the management capability usually encapsulates
functions that are specific to a management task, such as the aggregation and
upward propagation of a key performance indicator, deploying the
components of a single management capability is key for allowing various
kinds of tradeoffs, such as performance versus memory space. An example is
that of a virtualized environment, in which a management capability's
algorithm component is located in the runtime environment of a
programmable NIC due to performance reasons, while the information
retrieval and objective enforcement components are located in the virtual
hypervisor that is located “closer” to the virtualization system's policy
engines.
4 Scenario: OpenFlow Control and In-Network Management
We now describe an example scenario that combines some of the features
of centralized, hierarchical, distributed, and flat technology and show how
management capabilities facilitate the transition in parts of that scenario.
4.1 Application to Converged Systems
We consider how to apply management capabilities to a typical converged
system, in this case, IT and network convergence occurring for example in
large data centers. Fig. 4 illustrates how management capabilities can be
applied to an IT/network converged management and control system in
general. In the figure, both hierarchical and distribution concepts are indicated
and some of the management capabilities translate between these two
directions.