90 likes | 102 Views
This system efficiently manages, schedules, and monitors Data Placement activities in a Grid environment, making intelligent decisions based on job requirements. It integrates with various middleware and storage systems, ensuring reliability and recoverability from failures. The user-friendly Data Placement Scheduler offers a single interface for streamlined operations. Supported by multiple protocols, it simplifies data transfer and enhances job scheduling productivity.
E N D
Managing and Scheduling Data Placement (DaP) Requests in GRID Tevfik Kosar & Miron Livny
Motivation An intelligent data scheduling system is needed, which: • integrates/interacts with all different middleware and end-storage systems in the Grid • makes intelligent scheduling decisions depending on job requirements • is recoverable from storage system and network failures • easy to use, single interface
Data Placement Scheduler (DaPS) • Intelligently & reliably manages, schedules and monitors Data Placement (DaP) activities in a heterogeneous Grid environment • What Condor means for computational jobs, DaPS means the same for DaP jobs • Just submit a bunch of DaP jobs and then relax..
DaPS Client DaPS Client DaPS Client SRB Server SRM Server Sched. Exec. Accept DAPS Server Local Disk Req. Req. Req. Req. Req. Req. GridFTP Server GridFTP Server NeST Server DaPS Architecture Remote Local Get Queue Buffer Put Thirdparty transfer
DaP Job ClassAds [ Type = Reserve; Server = nest://turkey.cs.wisc.edu; Size = 100MB; reservation_no = 1; ] [ Type = Transfer; Src_url = srb://ghidorac.sdsc.edu/kosart.condor/x.dat; Dst_url = nest://turkey.cs.wisc.edu/kosart/x.dat; ] [ Type = Release; Server = nest://turkey.cs.wisc.edu; reservation_no = 1; ]
Supported Protocols • Currently supported: • FTP • GridFTP • NeST (chirp) • SRB (Storage Resource Broker) • SRM (Storage Resource Manager) • Very soon: • GDMP (Grid Data Management Pilot)
Reserve In & out Transfer in Job B Release in Transfer out Release out Case Study: DAGMan Add DaP jobs to the DAG structure PRE Job B POST
New DAGMan Architecture .dag File DAGMan DAGMan A DaPS Job Queue Condor Job Queue X X A C B Y D
Future Work • More intelligent scheduling • Data Level Management instead of File Level Management • Enhanced interaction with DAGMan • Possible integration with Kangaroo to keep the network pipeline full