Professional Documents
Culture Documents
A Functional Model of Cooperative System Management: by Bharat Bhushan and Ahmed Patel
A Functional Model of Cooperative System Management: by Bharat Bhushan and Ahmed Patel
A Functional Model of
Cooperative System
Management
There is a need for information exchange and
communication between management applications that are
distributed over a large system. In this situation, a
fundamental problem is how the management systems
should cooperate in order to achieve overall management
of a system. In order to address this problem, this article
presents a functional model of cooperative system
management. 1998 John Wiley & Sons, Ltd.
By Bharat Bhushan and Ahmed Patel*
Introduction
he concept of cooperative working
involves a number of persons working
on a common task. In computer supported cooperative work (CSCW), computers are used to support the groups of person
or application software working on common task.
CCC 10557148/98/03017011$17.50
Requirements of Cooperative
System Management
There are three main requirements for cooperative system management. The first is that of overall
system-wide management. When a managed system grows in size, the management functionality
171
needs to distribute for remote monitoring and control. Operation management and administration
must be done across the boundaries of a managed
system. System management should be proactive
and automatic and intelligence-based management techniques should be used.
The second requirement is the use of a system
management framework which provides a broad
range of management services. The system management model should support addition of new
management services to the managing system and
new managed resources to the managed system.
The system management model should also support interoperability among different types of
management applications.
Finally, cooperation among several management applications is required. Management applications are required to share management information and services. Cooperation system
management itself is an unorganized activity
because operations occur in real-time in an open
system and unexpected conditions may arise frequently. Also, information and management services required by system administrators and operators are incomplete and may contain only partial
information for day-to-day operations. Means of
coordination during cooperation are required in
order to organise the cooperative system management activity. Management applications should be
provided with means of coordination during
cooperation.
172
Four common operations for cooperative system management: send and receive to send and
receive messages and open and close end-point
for communication and service requests from
one managing module to the other.
173
Figure 1. The structure and information flow within the Functional Model.
174
Managing Modules
A managing module is an autonomous set of
management applications that provide services
defined in the OSI management functional areas.
Management applications that provide one type of
services are grouped into one managing module.
A managing module therefore can be considered
as a set of responsibilities delegated to its management applications.
A managing module is a set of objects, which
provide management services. The managing module is the base class for different types of managing modules, which represent various management functional areas. The managing module base
class represents the top level of management function hierarchy. As the functional model is
extended, other object classes can be inherited
from the base object class. Each management
application of a managing module can further
derive its own object class, resulting in a hierarchical managing module class structure.
For example, two managing modules, which are
derived from managing module base object class,
are the fault managing module and performance
managing module. These two managing modules
are classes within themselves. However, in the
hierarchy of co-operating managing modules, they
are derived instances of the base class which reside
a level above in the class hierarchy. The fault-fix
and fault-report applications of the fault module, which provide fault management services, are
grouped into the fault module (similar grouping
mechanism are described in references 3 and 4).
The calculate-throughput and test-capacity
applications, which provide performance management services, are grouped into the performance
module.
In order to achieve interoperability in this
scheme, a managing module class that is derived
from the base class can be given a unique name.
The names of other classes that are derived from
the base class can be formed by linking the names
of all its parent classes.
During cooperation, a managing module may
use the services that are provided by other
module(s). Also, management application in one
managing module may use services of other management applications of the same managing module.
Creating groups of management applications
Messages
Managing modules use messages to communicate with each other. Messages are pre-defined
and are exchanged in a real-time manner to carry
management
data,
management
operation
requests and operation results. Messages are structured in such a way that a message can carry an
operation request as well as a result (or
acknowledgement) of the message received.
New messages may also be introduced if the
functional model is to be extended. In order to
support this requirement, messages may also be
categorized on the basis of their purpose and such
categories can be represented by object classes.
The communications between managing modules
can be abstracted to form categories. For example,
an abstraction can be done on the basis of what
message is sent to which module. Other categories
of message may be acknowledgement type, event
report type and management information type.
This categorization can provide three classes of
messages from which new messages of a particular type can be derived.
Operations
Managing modules use a set of operations to
obtain, manipulate and share management information. Operations describe the action that a managing module takes when it receives a message.
Operations can be locally or remotely executable.
An operation can be characterized5 by the managing module that uses the operation, the rules that
are used in the execution of the operation, and the
events that invoke the operation. There are four
types of operations (see operations in Figure 1). The
managing modules use two types of operations for
cooperation and two types are used by management applications for system management.
Managing modules cooperate using operations
Call(management-application-name) and messageexchange
operations
send(message)
and
receive(message). A managing module uses
Call(management-application-name) when it wants to
175
provide a management service to another managing module. The send(message) and receive(message)
operations are used by managing modules to
exchange messages with each other.
Management applications use three system management operations, namely Get, Set and EventReport. Management applications use connection
management operations open-end-point(entityname) and close-end-point(entity-name to create and
close end-points for communication with system
management agents. When communicating with a
managing module, the entity-name contains the
logical address of a managing module. When communicating with an agent, the entity-name contains the logical name of an agent.
Relationship
The relationship (shown in Figure 2) is one of
the global components of the functional model and
is derived from the functional dependency6 of
managing modules. There can be many types of
relationship between managing modules. The
relationship relates managing modules by deciding three aspects of cooperation. The first concerns
which managing module uses services and management information of which managing module.
The second aspect concerns when they should
cooperate. The third concerns which of the related
managing modules initiates the cooperation.
Inserting and deleting links between object classes
can easily change the relationship among managing modules and new relationships can be established.
An example of a relationship, the uses relationship between the fault and the performance
module, is shown in Figure 2. This relationship
specifies that the performance module use the
fault report and fault fix services of the fault
module. By establishing this relationship, it is
explicitly stated that out of many managing modules it is the fault module that contacts the performance module. The relationship uses between
the two modules is a one-to-one relationship where
the performance module uses services provided
by only one managing module, the fault module.
176
Rules
Rules are also global parts of the functional
model and coordinate the messages exchanged
between managing modules. They specify what
operations a managing module should perform
when it receives a message or a predefined event
occurs. All managing modules follow a common
set of rules.
Rules bind messages and events with one and
more operations of a managing module. Constructing rules in this manner is also explained in
references 7 and 8. Rules are defined in such a way
that all possible message-exchanges between all
the components can be pre-specified. A rule is
specified in the following format:
if
(receive(message) ORAND event)
then
(operation(s) )
Communication Support
In order to allow managing modules to communicate with each other and with agents, the
functional model includes communication support. The communication support organizes the
communication paradigm and services that are
required for cooperative system management. The
messages and system management operations are
submitted to the communication support, which
carries them to managing modules or to the
agents. There are two types of communication that
are supported in the functional model: synchronous and asynchronous. With the synchronous
communication support, managing modules can
send Get and Set system management operations
to a system management agent. With the asynchronous communication support, managing
modules can share management information, send
management requests to each other, and receive
event notifications from agents. Cooperation and
system management require guaranteed delivery
of messages and system management operations.
In order to meet this requirement, the communication support provides reliable message exchange
between two modules.
177
178
Initialization Phase
The fault and performance modules start as
autonomous modules at state 1 and remain so
until at state 2. During the initialization phase, the
fault module is prompted by a message from an
agent. At state 2, a fault occurs and fault-report
receives a notification from an agent.
Cooperation Phase
Rule 1:
if (receive (resource status = 1) )
then
send (fault occurred)
179
Rule 2:
if (receive(fault occurred))
then
send (restart resource)
Rule 6:
if (receive (restart failed)
then
send (message accepted)
Rule 3:
if (receive(restart resource))
then
return value = call (fault-fix)
Termination Phase
In the termination phase, the decision to terminate can be taken by any of the managing modules
that cooperated. The reason for terminating the
cooperation is made clear by sending an appropriate message. The receiver of the message that terminates the cooperation must send an acknowledgement.
For
example,
the
fault-fix
management may not fix the fault. If the faultfix application informs the fault module that it
could not fix the fault, which it does by returning
the value 1, the following rule is used:
Rule 5:
if (return value = 1)
then
send (restart failed)
If the performance module receives restart failed, it sends message accepted to the fault module using rule 6.
180
Message/Events
From
To
resource status = 1
fault occurred
restart resource
resource restarted
restart failed
message accepted
Agent
fault module
performance module
fault module
performance module
performance module
fault module
fault module
performance module
fault module
fault module
performance module
Rule
Rule
Rule
Rule
Rule
Rule
Rule
1
2
3
4
5
6
Table 1. Rules and messages used to govern activity of fault and performance modules
decisions taken by a managing module are
conveyed to a management application with
the call(management-application-name) operation. In turn, the management application
called carries out the operation management.
In order to inform the managing modules of
the results of the operation, management
applications communicate with the managing
module using rules and relationship.
Group communication for distributed system
management: The AMIGO Activity Model was
used for the specification of participants and
activities in cooperative system management.
Coordination: Cooperative system management requires that a set of predefined guidelines to be used in order to coordinate the
cooperation between the managing modules.
The functional model achieves the goal of
coordination because the managing modules
are responsive to each other. They operate
simultaneously on a common task and share
management information and services. The
co-ordination mechanism used in the functional model makes use of a single set of predefined rules used by both managing modules. The need for cooperation is stated in
terms of a relationship between the managing modules.
Application Areas
There are two areas to which the functional
model can be readily applied: TMN for services
and networks management12,13 and intelligent
management agents.14
Todays telecommunication networks require
automation and enhanced control capabilities for
management and interfunctional area management is needed to support this requirement. The
TMN framework aims at the management of heterogeneous networks, services and equipment distributed over many operators and service providers. It provides rich management functionality
and has provision for interworking among multiple management and operational systems.15
Many types of relationship exist among the users
and the providers of telecommunication services
and networks. In this development, the idea of
cooperation among several management systems
is competent. In cooperation, many administrators
and operators share management information and
services to manage heterogeneous resources (e.g.
network elements) and operations. Therefore the
functional model can be used in the area of TMN.
References
1. B. Bhushan and A. Patel, Requirements and the concept of co-operative system management, International Journal of Network Management, 8 (3), 139
158, May-June 1998.
2. T. Danielsen, AAMThe AMIGO Activity Model,
in Computer Based Group Communication: The AMIGO
activity model, edited by Pankoke-Babatz, pp. 7297,
Ellis Horwood, Chichester, 1989.
3. S. M. Dauber, Finding fault, BYTE, p. 207, Section
State of the Art, March 1991.
4. D. Gaiti, Introducing intelligence in distributed systems management, Computer Communication, 17,
No. 10, October 1994.
5. P. Hennessy et al., Distributed work management:
activity coordination within the EuroCoOp Project,
Special Issue: Computer Supported Co-operative Work,
Computer Communication, 15, No. 8, 47788,
October 1992.
6. S. Ram, Deriving functional dependencies from the
entity-relationship model, Communications of the
ACM, 38, No. 9, September 1995.
7. J. Bowers et al., Local and global structuring of computer mediated communication: developing linguistic perspectives on CSCW in COSMOS, Proceedings
of Conference on Computer-Supported Co-operative
Work, September 1988, pp. 12539, Association for
Computing Machinery, Inc., 1988.
8. V. Tschammer et al., Co-operative management in
open distributed system, Computer Communication,
17, No. 10, October 1994.
9. S. Benford, Requirements of activities management,
Proceedings of the First European Conference on Computer Supported Co-operative WorkEC-CSCW 89,
1315 September 1989, pp. 27685, Elsevier Science,
London, 1989.
10. H. T. Smith, et al., The activity model environment:
11.
12.
13.
14.
15.
16.
17.
18.
181
an object-oriented framework for describing organisational communication, Proceedings of the First European Conference on Computer Supported Co-operative
WorkEC-CSCW 89, 1315 September 1989,
pp. 16073, Elsevier Science, London, 1989.
B. Bhushan, Application of Co-operative Working to the
Management of a Heterogeneous Distributed System,
PhD thesis dissertation, Department of Computer
Science, University College Dublin, Belfield, Dublin
4, Ireland, February 1997.
S. R. Hedberg, Industry Spotlight: The telecommunication network of the new millennium, IEEE Parallel
and Distributed Technology, 4, No. 1, Spring 1996.
S. R. Hedberg, AIs impact in telecommunications
today and tomorrow, IEEE Expert, 11, No. 1, February 1996.
A. M. Hayashi and S. E. Varney, Six hot technologies
for the 21th century: software agents, Datamation,
August 1996.
CCITT Recommendation M.3010. Principles for a
Telecommunication Management Network (TMN), ITUT, Geneva, 1989.
D. Weld, The role of intelligent systems in the
national information infrastructure, AI Magazine,
4564, March 1995.
C. Hsinchun et al., Toward intelligent meeting
agents, IEEE Computers, August 1996.
B. F. Lubomir et al., Distributed computing using
autonomous objects, IEEE Computers, August 1996.