Information Technology Reference
In-Depth Information
formed according to measures of context and trust and tailored to the individual pref-
erences and circumstances of users. Flexible techniques will be needed to filter infor-
mation and perform access control, as well as defining and enforcing privacy. Users
will expect management functions to be invisible and carried out automatically.
We are developing a policy-based architecture that supports management at vary-
ing levels of granularity, using the concept of a self-managed cell (or simply a cell).
A cell consists of a set of hardware and software components that represent an admin-
istrative domain. Cells are able to function autonomously and thus capable of self-
management. A cell could represent the resources available in a PDA, a body area
network of physiological sensors and controllers. At the enterprise level, a cell could
represent the resources and application components relating to a set of collaborating
partners forming a virtual organisation spanning multiple countries. In each case, cells
include and evolve the required management services, appropriate to the scale and
environment of the cell. These management services interact with each other through
asynchronous events exchanged over an event bus. In essence, a cell is a “closed-
loop” system where changes of state in the managed objects and resources trigger
adaptation that in turn affects the state of the system. In ubiquitous environments, the
cells would also typically include management components that provide service dis-
covery and contextual management.
A cell includes a policy-driven agent that supports context-based and trust-based
access control and system adaptation for one or more ubiquitous devices. Cells can
load additional management functions and organise themselves into larger manage-
ment cells through federation and nesting. Potentially, each ubiquitous device that a
user carries, and each device situated in the environment, is capable of being a self-
managed cell and running a management agent that carries out management functions
and policies. In practice, we envisage that some devices (e.g. sensors) will be too
primitive to run their own management agent, but will be capable of being managed
by an external cell, such as a mobile phone, over a wireless link, such as bluetooth.
This extended abstract introduces the architecture of self-managed cells.
2 Self-managed Cells
Each self-managed cell consists of a number of core management components: the
cell watchdog, the event service, the discovery service, the policy service, and the
domain service. Cells can also load components for context and trust management as
well as monitoring and intrusion detection. Proxies are required to interact with the
various communication interfaces of devices and managed components, for example
to enable cell policies to perform actions on device-specific management interfaces,
and to convert low-level signals to cell events. The following outlines the core ser-
vices of each self-managed cell.
2.1 Cell Watchdog
When a cell is first instantiated, it starts up the cell watchdog. This is a special ser-
vice that is responsible for loading and instantiating the core components of the cell,
typically from local storage (e.g. a memory card), or from a remote cell. The cell
