AMUSE
Autonomic Management of Ubiquitous
Systems for e-Health
Prof. J. Sventek
University of Glasgow
joe@dcs.gla.ac.uk
In collaboration with M. Sloman, E. Lupu, and N.
Dulay of Imperial College London
Executive Summary
• Increasing complexity of distributed application
systems leads customers to desire automated
management of such systems.
• Work at Agilent/Glasgow has yielded an architectural
pattern and prototype implementations for closedloop management of distributed application systems.
• Imperial has established itself as one of the premier
research groups for policy-based management.
• AMUSE is focused on integrating these
complementary competencies to address automated
management of e-Health applications
What is closed-loop
• Systems that utilize feedback are called closed-loop
control systems
• The feedback is used to make decisions about
changes to the control signal that drives the system
Closed-loop Management
Pattern (Self-Managed Cell)
Management
Application
Measurement
Raw
Measurement
Analysis,
Simulation,
Optimization
Provisioning
Trends &
Prediction
Event Bus
Policy
Management
Measurement
Adapters
Service Goals
System Policy
“System”
Configuration
Topology,
Other
“System” Under Test
Two-level
nesting
Measurement
Management
Application
Level n
Analysis,
Simulation,
Optimization
Raw
Measurement
Provisioning
Trends &
Prediction
Event Bus
Measurement
Adapter
Policy
Management
Service Goals
System Policy
Meas
Infer
Level n-1
Prov
Event Bus
Level
n-2
Policy
Agents
“System”
Config
“System”
Configuration
Topology,
Other
Policy-based Management
• Traditional management systems are imperative – i.e.
the manager explicitly programs actions to take
• Policy-based management systems are declarative –
i.e. the manager indicates what outcomes are desired,
and the management system components attempt to
reach these goals
• All aspects of correct system behaviour are covered
by policies – e.g. security, performance, fault
handling, configuration & scale, etc.
Research Issues Addressed by
AMUSE
•
•
•
•
•
•
•
•
How to specify the required management functionality of a SMC, dynamically
add/remove resources and management services into SMCs, instantiate an SMC
and deploy its components across distributed nodes within a network or
application?
How to design adaptive and context-aware SMCs? What consistency and
integrity constraints must be preserved within an SMC? How to express, deploy
and enforce policies in SMCs?
What information models are necessary for managed resources and the
management services - including state information, performance attributes and
events? How to negotiate which events are propagated?
How to ensure that the SMC paradigm scales from SMCs with limited resources
(e.g. a portable body area network) to SMCs with large resources (e.g. for
managing an application distributed across server clusters)?
What management interactions are needed for composed, layered and peer-topeer SMC structures? When to export/hide SMC management functionality?
How to refine policies from composite SMCs to policies for encapsulated SMCs,
particularly when the nested SMCs enter or leave the enclosing one dynamically?
How to support dynamic adaptation and configuration of SMCs into larger SMC
infrastructures? What policies and constraints are needed to manage this?
What aspects of management are context-dependent?
AMUSE Work Packages
i.
The specification of the generic and extensible
SMCs. Within an SMC, the primary areas of
research are the core services for
Interaction/Adaptation, Policy, Context, and
Measurement/Control.
ii. Investigations into the federation of SMCs, the
layering of SMCs, and their integration with legacy
management systems/technologies.
iii. Investigations into the composition of SMCs within a
single administrative domain,
iv. The development of two e-Health prototypes to
validate the SMC architecture.