Parts | Days | Selection | Search | Downloads | Help
T: Teilchenphysik
T 509: Grid II
T 509.4: Talk
Thursday, March 30, 2006, 17:10–17:25, HG2-HS7
Job and Data Co-Scheduling in Data Intensive HEP Applications — •Michael Ernst1, Patrick Fuhrmann1, and Ramin Yahyapour2 — 1DESY, Notkestr. 85, 22607 Hamburg — 2Universitaet Dortmund, Otto-Hahn-Str. 4, 44227 Dortmund
The Grid is designed to offer transparent access to resources of very different nature as for instance CPU, network, data or software.
Grid scheduling and brokerage needs the ability to orchestrate the availability of all required Grid resources to execute Grid jobs. A complex negotiation process is necessary that combines the access policies and local resource management of the individual resources and their owners.
Addressing the data intensive applications in HEP a Grid Storage Element (SE) has been developed consisting of dCache as the core storage system and an implementation of the Storage Resource Manager (SRM). This SE allows both local (POSIX-like) and Grid access (GridFTP) to mass storage facilities based on hierarchies of tape and disk technology as well as small to large disk-only configurations. The Storage Resource Manager (SRM) protocol, supporting secure data transfers with protocol negotiation and reliable replication mechanisms over wide area networks, has become a standard for Grid interfaces to managed storage.
In this talk we will propose an extended scheduling architecture, which adds a scheduling framework on top of existing compute and storage elements in order to improve coordination between data and workload management. Recent results of the development work that have been achieved so far will be presented.