On-demand Provisioning of Workflow Middleware and s An Overview University of Stuttgart Universitätsstr. 8 70569 Stuttgart Germany Karolina Vukojevic-Haupt, Florian Haupt, and Frank Leymann Institute of Architecture of Application Systems karolina.vukojevic@iaas.uni-stuttgart.de Phone +49-7-685 88484 Fax +49-7-685 8847
Agenda Motivation Approach for ODP Extended Classification of Binding Strategies Selection and Package Selection Process Comparison of Binding in SOC and ODP Karolina Vukojevic-Haupt, IAAS On-demand Provisioning of Workflow Middleware and s
Motivation Basic assumption in Oriented Computing (SOC) s always on and available Suitable for production workflows in the business domain s are typically used continuously There are domains where services are used rarely and not regularly E.g. simulation workflows in the escience domain Keeping services always on and available is a waste of resources Our approach to solve this problem: On-demand provisioning of workflow execution middleware and services (ODP) Karolina Vukojevic-Haupt, IAAS On-demand Provisioning of Workflow Middleware and s
On-Demand Provisioning SERVICE SERVICE 7 S 7 S DB AS Audit DB AS WORKFLOW MIDDLEWARE Workflow Engine Enterprise Bus SERVICE S Provisioning Engine DB AS Karolina Vukojevic-Haupt, IAAS On-demand Provisioning of Workflow Middleware and s 4
On-Demand Provisioning WORKFLOW MIDDLEWARE Workflow Engine SERVICE Users can run their simulation workflows S DB in AS the Cloud with only one click Audit Provisioning Engine Optimizing resource allocation DB AS SERVICE 7 Only modeling tool and bootware Enterprise local on user s machine SERVICE Bus Middleware and services are provisioned on demand S Exploiting cloud characteristics DB Reuse of existing provisioning technologies Installing and running the workflow middleware and services is handled automatically and invisible in the background S 7 AS Karolina Vukojevic-Haupt, IAAS On-demand Provisioning of Workflow Middleware and s 5
Topology ESB Classification of Binding Strategies Workflow Legend deployed in advance deployed on demand A B C D static binding Discovery Selection Deployment Logic dynamic dynamic binding (EP X ) binding (EP Y ) deploy Provisioning Logic dynamic binding (EP Z ) 4 provision Infrastructure and Middleware EP Y EP Z deploy Implementation S Implementation S 4 relation connected to S static binding S dynamic binding Middleware and Infrastructure dynamic binding with service deployment Application Server Database relation hosted on Operating System dynamic binding with software stack provisioning Virtual Machine Vukojevic-Haupt, K.; Karastoyanova, D.; Leymann, F.: On-demand Provisioning of Infrastructure, Middleware and s for Simulation Workflows. In: Proceedings of SOCA 0 Karolina Vukojevic-Haupt, IAAS On-demand Provisioning of Workflow Middleware and s 6
Topology ESB Classification of Binding Strategies Workflow Legend deployed in advance deployed on demand A B C D static binding Discovery Selection Deployment Logic dynamic dynamic binding (EP X ) binding (EP Y ) deploy Provisioning Logic dynamic binding (EP Z ) 4 provision Infrastructure and Middleware EP Y EP Z deploy Implementation S Implementation S 4 relation connected to S static binding S dynamic binding Middleware and Infrastructure dynamic binding with service deployment Application Server Database relation hosted on Operating System dynamic binding with software stack provisioning Virtual Machine Vukojevic-Haupt, K.; Karastoyanova, D.; Leymann, F.: On-demand Provisioning of Infrastructure, Middleware and s for Simulation Workflows. In: Proceedings of SOCA 0 Karolina Vukojevic-Haupt, IAAS On-demand Provisioning of Workflow Middleware and s 7
Selection and Package Selection Process Registry Package Repository Workflow Engine N Discovery 4 5 N Selection 6 PC 8 Package Selection 9 service packages service call activity Enterprise Bus 7b Provisioning Manager SPT, Cloud a SPT, Cloud b SPT, Cloud b 7a forward 0 Legend functional requirements N traditional non-functional requirements provisioning requirements PC provisioning capabilities ccc S x provisioned in advance forward Implementation S y Middleware and Infrastructure provisioned on demand SPT, Cloud a provision engine is able to provision service packages Vukojevic-Haupt, K., Haupt, F., Karastoyanova, D., Leymann, F., (04). Selection for On-demand Provisioned s. of type SPT in the Cloud a In: Proceedings of the 8th IEEE International EDOC Conference (EDOC 04) Karolina Vukojevic-Haupt, IAAS On-demand Provisioning of Workflow Middleware and s 8
Separation of Concerns Registry Package Repository Workflow Engine N Discovery 4 5 N Selection 6 PC 8 Package Selection 9 service packages create FEM parameters Enterprise Bus 7b Provisioning Manager SPT, Cloud a SPT, Cloud b SPT, Cloud b Legend functional requirements N traditional non-functional requirements provisioning requirements PC provisioning capabilities ccc S x 7a forward provisioned in advance forward ESB & Registry Implementation S y Traditional Selection and Routing Middleware and Infrastructure 0 provision service package of S y a) provision infrastructure and middleware b) deploy service implementation provisioned on demand SPT, Cloud a provision engine is able to provision service packages of type SPT in the Cloud a Karolina Vukojevic-Haupt, IAAS On-demand Provisioning of Workflow Middleware and s 9
Separation of Concerns Provisioning Manager & Package Repository Provisioning related functionality Package Selection Provisioning of services Workflow Engine N Discovery Registry 4 5 N Selection 6 PC Package Repository 8 Package Selection 9 service packages Legend functional requirements N traditional non-functional requirements provisioning requirements PC provisioning capabilities ccc create FEM parameters Designed to be extensible Support multiple Clouds Support multiple service package formats Enables reuse of existing S service packages available x in the web 7a Enterprise Bus forward provisioned in advance 7b forward Provisioning Manager Implementation S y Middleware and Infrastructure SPT, Cloud a SPT, Cloud b SPT, Cloud b 0 provision service package of S y a) provision infrastructure and middleware b) deploy service implementation provisioned on demand SPT, Cloud a provision engine is able to provision service packages of type SPT in the Cloud a Karolina Vukojevic-Haupt, IAAS On-demand Provisioning of Workflow Middleware and s 0
Binding in SOC and ODP: Publishing a SOC register service ( FC + NFC + EP ) Registry Provider ODP register service package ( PC ) Package Repository Package Provider register service ( FC + NFC + S ) Registry register service ( EP ) Provider t 0 t n, n>0 LEGEND FC = functional capabilities NFC = non-functional capabilities EP = endpoint PC = provisioning capabilities S = service package reference Karolina Vukojevic-Haupt, IAAS On-demand Provisioning of Workflow Middleware and s
Binding in SOC and ODP: Calling a SOC N 0 initial provisioning in advance Enterprise Bus S x Consumer 4 service response service response Provider ODP N provision on demand 6 deprovision Consumer 5 service response ODP Middleware 4 service response Implementation S y Middleware and Infrastructure provisioned on demand LEGEND = functional requirements N = non-functional requirements = provisioning requirements Karolina Vukojevic-Haupt, IAAS On-demand Provisioning of Workflow Middleware and s
Comparison of SOC and ODP: Publish Find Bind SOC ODP Publish Functional capabilities Non-functional capabilities Endpoint Functional capabilities Non-functional capabilities package reference Find selection package selection Bind Bind to endpoint Provision service Bind to endpoint Karolina Vukojevic-Haupt, IAAS On-demand Provisioning of Workflow Middleware and s
Summary Always on semantics of SOC is not for services that are used rarely and irregularly Our approach: On-Demand Provisioning (ODP) Users can run their workflows in the cloud with only one click Complexity is hidden in the middleware (extended ESB) Exploits cloud characteristics & uses existing provisioning technologies Introduces additional challenges More ODP-related topics are discussed in the paper And there is also a poster! Contact Karolina Vukojevic-Haupt karolina.vukojevic@iaas.uni-stuttgart.de Karolina Vukojevic-Haupt, IAAS On-demand Provisioning of Workflow Middleware and s 4
Thank you 5