An Aspect-Oriented Approach - Autonomics Development: A Domain-Specific Aspect Language Approach - page 50

Table 3.3. Four aspects of self-management with autonomic

computing.

Concept Current Computing Autonomic Computing

Self- Corporate data centres have Automated configuration of

configuration multiple vendors and plat- components and systems fol-

forms. Installing, configuring, lows high-level policies. Rest

and integrating systems is of system adjusts automati-

time consuming and error cally and seamlessly.

prone.

Self- Systems have hundreds of Components and systems

optimisation manually set, nonlinear tun- continually seek opportuni-

ing parameters, and their ties to improve their own per-

number increases with each formance and eciency.

release.

Self-healing Problem determination in System automatically detects,

large, complex systems can diagnoses, and repairs local-

take a team of programmers ised software and hardware

weeks. problems.

Self- Detection of and recovery System automatically defends

protection from attacks and cascading against malicious attacks or

failures is manual. cascading failures. It uses early

warning to anticipate and
prevent systemwide failures.

3.7 AOP and Autonomics
Autonomic computing is an initiative proposed by the IBM corporation to over-
come an impending complexity crisis in the development of applications [52]. Ac-
cording to IBM, this complexity is growing beyond human ability to manage it.
To overcome this, IBM proposes that systems are developed to manage them-
selves, given high-level objectives by administrators, so that they may adjust their
operation, workloads, demands and external conditions in the face of hardware or
software failures. IBM cites four aspects of self-management, which are detailed
in Table 3.3 [56].
To meet the autonomic computing vision of self-managing, self-healing and
self-optimising systems requires a system to be able to dynamically adapt to its
environment. However, a key challenge limiting the use of autonomic features in
applications is the lack of tools and frameworks that can alleviate the complexities
stemming from the use of manual development methods [56].
McKinley et al. [69] define two general approaches to implement adaptive
software: