Optimizing performance of workflow executions under authorization control

Preview

Text WRAP_THESIS_Chaudhary_2013.pdf - Submitted Version Download (14MB) | Preview

Request Changes to record.

Abstract

“Business processes or workflows are often used to
model enterprise or scientific applications. It has
received considerable attention to automate workflow
executions on computing resources. However, many
workflow scenarios still involve human activities and
consist of a mixture of human tasks and computing
tasks.
Human involvement introduces security and
authorization concerns, requiring restrictions on who
is allowed to perform which tasks at what time. Role-
Based Access Control (RBAC) is a popular authorization
mechanism. In RBAC, the authorization concepts such as
roles and permissions are defined, and various
authorization constraints are supported, including
separation of duty, temporal constraints, etc. Under
RBAC, users are assigned to certain roles, while the
roles are associated with prescribed permissions.
When we assess resource capacities, or evaluate the
performance of workflow executions on supporting
platforms, it is often assumed that when a task is
allocated to a resource, the resource will accept the
task and start the execution once a processor becomes available. However, when the authorization policies
are taken into account,” this assumption may not be
true and the situation becomes more complex. For
example, when a task arrives, a valid and activated
role has to be assigned to a task before the task can
start execution. The deployed authorization
constraints may delay the workflow execution due to
the roles’ availability, or other restrictions on the
role assignments, which will consequently have
negative impact on application performance.
When the authorization constraints are present to
restrict the workflow executions, it entails new
research issues that have not been studied yet in
conventional workflow management. This thesis aims to
investigate these new research issues.
First, it is important to know whether a feasible
authorization solution can be found to enable the
executions of all tasks in a workflow, i.e., check the
feasibility of the deployed authorization constraints.
This thesis studies the issue of the feasibility
checking and models the feasibility checking problem
as a constraints satisfaction problem.
Second, it is useful to know when the performance of
workflow executions will not be affected by the given
authorization constraints. This thesis proposes the
methods to determine the time durations when the given
authorization constraints do not have impact.
Third, when the authorization constraints do have
the performance impact, how can we quantitatively
analyse and determine the impact? When there are multiple choices to assign the roles to the tasks,
will different choices lead to the different
performance impact? If so, can we find an optimal way
to conduct the task-role assignments so that the
performance impact is minimized? This thesis proposes
the method to analyze the delay caused by the
authorization constraints if the workflow arrives
beyond the non-impact time duration calculated above.
Through the analysis of the delay, we realize that the
authorization method, i.e., the method to select the
roles to assign to the tasks affects the length of the
delay caused by the authorization constraints. Based
on this finding, we propose an optimal authorization
method, called the Global Authorization Aware (GAA)
method.
Fourth, a key reason why authorization constraints
may have impact on performance is because the
authorization control directs the tasks to some
particular roles. Then how to determine the level of
workload directed to each role given a set of
authorization constraints? This thesis conducts the
theoretical analysis about how the authorization
constraints direct the workload to the roles, and
proposes the methods to calculate the arriving rate of
the requests directed to each role under the role,
temporal and cardinality constraints.
Finally, the amount of resources allocated to
support each individual role may have impact on the
execution performance of the workflows. Therefore, it
is desired to develop the strategies to determine the
adequate amount of resources when the authorization
control is present in the system. This thesis presents the methods to allocate the appropriate quantity for
resources, including both human resources and
computing resources. Different features of human
resources and computing resources are taken into
account. For human resources, the objective is to
maximize the performance subject to the budgets to
hire the human resources, while for computing
resources, the strategy aims to allocate adequate
amount of computing resources to meet the QoS
requirements.

Request changes or add full text files to a record

Repository staff actions (login required)

View Item