Environmental Engineering Reference
In-Depth Information
“Engineer Autonomicity” has an implied Systems and/or Software
Engineering view, under which autonomic function would be engineered
into the individual systems. “Learn Autonomicity” has an implied AI, evo-
lutionary computing, and adaptive learning view, where the approach would
be to utilize algorithms and processes to achieve autonomic behavior. How-
ever, both approaches rely on each other in achieving the objectives set out
in Automatic Computing. Autonomic Computing may prove to require a
greater collaboration between the intelligence-systems research and system-
and software-engineering fields to achieve the envisaged level of adaptation
and self-management within the autonomic computing initiative.
8.1.3 Necessary Constructs
Considering these autonomic properties, the key constructs and principles
that constitute an Autonomic Environment are:
•
∗
•
AE
=
MC
+
AM
•
Selfware; Self-
Control Loop; Sensors+Effectors
•
AE ↔ AE
Selfware; Self-
:
The principle of selfware (self-managing software and firm-
ware) and the need for self-
∗
∗
properties were discussed in the previous
sections.
AE=MC+AM
:
Figure
8.2
represents a view of an architecture for an au-
tonomic element, which consists of the component to be managed and
the autonomic manager [
69
,
154
]. It is assumed that an autonomic man-
ager (AM) is responsible for a managed component (MC) within a self-
contained autonomic element (AE). This AM may be designed as part
of the component or may be provided externally to the component, as
an agent, for instance. Interaction will occur with remote AMs (e.g.,
through an autonomic communications channel) through virtual, peer-
to-peer, client-server [
11
], or grid [
33
] configurations.
Control Loop, Sensors+Effectors
:
At the heart of any autonomic system ar-
chitecture are sensors and effectors [
42
]. A control loop is created by
monitoring behavior through sensors, comparing this with expectations
(historical and current data, rules, and beliefs), planning what action is
necessary (if any), and then executing that action through effectors [
68
].
The control loop, a success of manufacturing science for many years, pro-
vides the basic backbone structure for each system component [
41
].
IBM represents this self-monitor-self-adjuster control loop as the monitor,
analyze, plan, and execute (MAPE) control loop. The monitor and analyze
parts of the structure process information from the sensors to provide both
self-awareness and an awareness of the external environment. The plan
and execute parts decide on the necessary self-management behavior that
will be executed through the effectors. The MAPE components use the
