Information Technology Reference
In-Depth Information
individual network elements or at one of the lower layers of a network
protocol stack, down to the hardware of individual network elements (e.g.
network switches).
In a typical management task, for instance, a low-level objective is violated
in a management capability and resolved by the next upstream capability
using a specific self-adaptation mechanism without the need to report to a
general management point. This example illustrates the ability of the proposed
framework to adopt self-management principles, which are considered as an
important means for local self-management operations in order to improve the
management/control systems overall performance and to avoid explicit
interactions with human beings.
3.4 Realization Options
Implementing MC-based network management and control works by defining
the logic of management processes first, and then deploying management and
control functions onto the system in question. The definition of management
processes can be done by a language that is runtime-independent, which
allows specifying functional elements (e.g. network elements), the functions
these elements are to carry out, and how communication between elements is
done.
In particular, the organization and collaboration functions may be
implemented using different kinds of programming/communication models.
Based on our assessment, REST (representational state transfer) and RPC
(remote procedure call) are suitable for organization and collaboration,
respectively. The reason is that organization tasks are likely to be more
abstract in that an objective is modeled as a resource and manipulated via one
of the few generic methods according to REST. For collaboration, it is
naturally the case that algorithms (e.g. for monitoring, fault detection,
anomaly detection) are rather diverse and require specific interfaces to
communicate with each other, specific to the algorithm. These may be more
suitable designed via RPC or a new protocol suit is defined for that.
Once a management process is defined, it is to be deployed. This phase
requires specification of how to deploy a management function, and available
support, e.g. in terms of compilers, will determine possible target runtime
environments (also hardware is possible). For instance, OSGi is considered a
suitable candidate to implement individual management functions, and brings
with it already great flexibility and system support (such as discovery of
services, which would be mapped to the management capabilities, etc.) that
allows to easily relocate functions. Moreover, it is possible that existing
hypervisor environments can be used as runtime environments in virtual
environments. From these, in a transitional step, similar capabilities can then
Search WWH ::




Custom Search