AppLeS / Network Weather Service IPG Pilot Project FY’98 Francine Berman U. C. San Diego and NPACI Rich Wolski U.C. San Diego, NPACI and U. of Tennessee This presentation will probably involve audience discussion, which will create action items. Use PowerPoint to keep track of these action items during your presentation In Slide Show, click on the right mouse button Select “Meeting Minder” Select the “Action Items” tab Type in action items as they come up Click OK to dismiss this box This will automatically create an Action Item slide at the end of your presentation with your points entered.

User’s Goal = Program Performance IPG resources + Globus services = “abstract machine” Applications must be able to achieve performance on IPG “abstract machine” Application scheduling promotes program performance

Scheduling makes a difference. Compile-time Blocked Partitioning Run-time AppLeS Partitioning

Scheduling and the IPG Usability, Integration development of basic IPG/Globus infrastructure Development of persistent IPG/Globus testbed Performance “IPG - aware” programming Short-termMedium-termLong-term Application scheduling Resource scheduling Throughput scheduling Multi-scheduling Resource economy Integration of schedulers and other tools, performance interfaces Experience with Pilot IPG Development of prototype large-scale applications Support for multiple user communities Development of necessary research

AppLeS = Application-Level Scheduler AppLeS incorporates –application-specific information –dynamic information –prediction Schedule developed to optimize user’s performance measure –minimal execution time –turnaround time = staging/waiting time + execution time –other measures: precision, resolution, speedup, etc. NWS (Wolski) User Prefs App Perf Model Planner Resource Selector Application Act. IPG /Globus infrastructure

NWS = Network Weather Service (Wolski) NWS is standalone project –used by PACIs, Labs and others for dynamic resource information –provides dynamic forecasts for AppLeS –can be used as part of IPG performance toolkit NWS –monitors current system state –provides best forecast of resource load from multiple models Sensor Interface Reporting Interface Forecaster Model

FY98 Pilot Project: INS2D AppLeS Building prototype AppLeS scheduler for INS2D that will become model for other parameter studies AppLeS INS2D scheduler –first phase targets on interactive UCSD cluster –goal is to minimize turnaround time for user turnaround time = wait time + execution time

INS2D AppLeS Architecture AppLeS schedules work on interactive UCSD cluster AppLeS tuned to leverage underlying resource management system –sockets now, Globus and batch/interactive platforms next –API allows expansion for other parameter study applications AppLe S API Resources App- specific case gen. Exp Act Sched. Act Exp

AppLeS/NWS Pilot Project FY98 AppLeS (Berman,Wolski) Design and deployment of INS2D AppLeS for interactive cluster at UCSD –focus on socket design –developing general API for wider application class Conduct experiments to assess performance of INS2D AppLeS on interactive UCSD cluster NWS (Wolski) Pilot project integrating NWS and Globus –NWS/MDS integration to be demonstrated at SC98 –Performance prediction of Globus services focus on GRAM startup time Deployment of NWS on interactive UCSD cluster to support INS2D AppLeS

Current Status AppLeS (Berman,Wolski) INS2D AppLeS designed, software being developed Comparison experiments being designed Plan to have some results for 11/15 NASA Performance Workshop paper deadline NWS (Wolski) NWS/MDS currently being integrated for SC98 demo Currently running NWS experiments to determine GRAM startup times as part of end-to-end application performance Deploying NWS on PACIs and other sites