A Workflow Mapping Mechanism for establishing Quality of Service Guarantees

1
A Workflow Mapping Mechanism for establishing
Quality of Service Guarantees

• Dimosthenis Kyriazis
• National Technical University of Athens
• Telecommunications Laboratory
• e-mail dkyr_at_telecom.ntua.gr

2
Workflow Models

• Abstract Workflow
• tasks are described in an abstract form without
  referring to specific Grid resources for task
  execution
• provides only service semantic information on how
  the workflow has been composed.
• Concrete Workflow
• tasks are bound to specific resources
• provides both service semantic and execution
  information on how the workflow has been composed.

3
Workflow Mapping with regard to QoS

• Since
• each workflow contains processes (also called
  service types) that can be executed from a set of
  service instances (candidates), which are
  annotated with QoS information.
• A method is needed to
• map workflow processes to Grid service instances
  with regard to the provided QoS metrics.
• Workflow Mapping Mechanisms
• estimate, calculate and conclude to the workflow
  mapping and the selection of the available
  service types and instances
• deliver end-to-end QoS across a federation of
  providers.

4
Workflow Mapping Mechanism Overview

• Input of the mechanism
• Abstract Workflow
• User Constraints preferences hard limits on
  the requirements and preferences - importance
  factors on the aforementioned requirements
• Service Providers QoS Information Each service
  instance is annotated with QoS information.

5
QoS Parameters

• A classification of the QoS parameters is stated
  as a direct consequence of their logical
  categorization
• User-defined Parameters requirements posed by
  the user who initiates a workflow execution
  process (e.g. maximum workflow cost).
• Application Parameters offered quality level
  from the application perspective (e.g.
  application configuration may be reflected to the
  execution availability of a task).
• Resource Parameters relate to all types of
  resources, including computational, storage and
  network resources.
• Initial parameters of our study
• Characterisation (expressed as percentage) and
  calculated as follows
• Cost (expressed as cost units)
• Time (expressed as time units)

6
Use Cases based on Constraints Preferences

• Based on the users constraints and preferences
  the following cases can be are identified
• Thresholds. Users set thresholds / restrictions
  to the values of one or more parameters.
• Specific Parameter Optimization. Users set a high
  preference on one of the parameters and therefore
  the workflow mapping is performed considering the
  specific parameter as the one of major
  importance. The output will be a concrete
  workflow achieving the maximum average
  characterization, minimum workflow cost or
  minimum workflow execution time.
• Optimum Solution. In this case the users do not
  set a preference and as a result the weight
  attributes for the parameters are equal. The
  algorithm execution results to a concrete
  workflow where the service instances selected
  offer the optimum value of availability and
  execution time for their corresponding cost.

7
Algorithm Description (1)

• The algorithm used for the workflow mapping
  consists of four major steps
• Calculation of auxiliary values based on the QoS
  parameters values exposed from each service
  instance. The reason of this step is to define
  metrics that can be used by the algorithm to
  characterize the level of QoS provided by the
  service instances and proceed with selections
  based on these metrics.
• Slope values enact how important each parameter
  is considered to be by the user and may get
  values 0 Slope 1.

8
Algorithm Description (2)

• Initial workflow mapping is performed by
  selecting instances that meet the users
  requirement for characterization without
  violating the cost constraint.
• Definition of a candidate for each workflow
  process based on the outcome of Step 1.
• Execution of an embedded scheduling algorithm
  that allows more than one service instances to be
  selected per service type in order to meet time
  constraints set by the user.

9
Experimental Results
Scenario User sets QoS constraints and requests
Time Optimization
10
Workflow Sharing

• For the cases where workflow engines share a
  workflow database / registry, workflow mapping
  can also be performed since the input of the
  algorithm is an abstract workflow. The latter
  provides the ability to the users to define
  workflows in a flexible way, isolating execution
  details thus sharing of workflow descriptions
  between Grid users is feasible.
• The Workflow Mapping Mechanism is implemented as
  a Web Service which allows the use of it as a
  stand-alone module in order to achieve the
  optimum selection of service instances from a
  given service type or from an abstract workflow,
  with regard to QoS information.

11
Vision (1)

• CHALLENGERS (CHALLENGEs in gRidS
  challengers-org.eu)
• Specific Support Action
• Investigates enabling technologies and market
  trends
• Consolidates and describes the Vision of the
  research community for the coming decade
• Will recommend a Research Agenda and a Roadmap of
  key technology challenges
• Regarding Workflows and their maturity level
• can be considered reasonably mature for certain
  process oriented applications in grid context
• technology with proven potential at the research
  level
• industrially limited technology with high
  research maturity
• technology with clear research goals, however,
  has to evolve towards SOKU or other new paradigms
• maturity estimation 3-5 years from now

12
Vision (2)

• Workflows and Grid Programming
• A Grid infrastructure will enalble the enacting
  of a coordinated set of services encapsulating
  and virtualizing resources. Workflows are only a
  facet of the complex and multifaceted approach to
  the Grid programming.
• There is also an interdependence between user
  interfaces and programmability such as workflows.
  
• Workflows and Composability
• Using workflows to compose services into the Grid
  applications of the future is fundamental to the
  future of the Grid. Pour research effort into the
  automatic composability of semantically-rich,
  highly-distributed, secure Grid services.
• Workflows and Service Discovery
• Directly related to automatic composability and
  the automatic creation of workflows to solve
  problems. The time to maturity is very
  aggressive.
• Workflows and SLA Management
• SLAs potentially provide a means by which
  separate services can automatically agree to work
  together as part of a workflow. They may be part
  of the overall solution to the automatic
  composability problem but they are only part of
  it.

13
Conclusions

• Workflow Mapping Mechanisms
• define a concrete workflow based on an abstract
  workflow and on various parameters (user and
  service provider side) having abstract models as
  input allows the sharing of workflows.
• establish Quality of Service Guarantees, which is
  expected to increase the effort to provide the
  Grid environment with a dynamic QoS capability.
• The future challenge is the way to express in
  quantitative terms the meaning of business
  relationships and business strategies such as
  strategic partnerships.
• The approach presented is part of the ongoing
  research effort in the framework of NextGRID IST
  Project.

14
Thank you!
Questions ?