Скачать презентацию Preparation for the Tera Grid Account Synchronization and Скачать презентацию Preparation for the Tera Grid Account Synchronization and

d42c5e3e58028599d92adae32345c23c.ppt

  • Количество слайдов: 8

Preparation for the Tera. Grid: Account Synchronization and High Performance Storage Bobby House Preparation for the Tera. Grid: Account Synchronization and High Performance Storage Bobby House

About the Tera. Grid About the Tera. Grid

Tera. Grid Integration q Integration Challenges Accounting Network Security Tera. Grid Software q Local Tera. Grid Integration q Integration Challenges Accounting Network Security Tera. Grid Software q Local Accounting System Interface Account Creation Tracking User Allocations

NCSA/TG Allocations User and project packets Frost (IBM BG/L) TGCDB FEN 1 DBST Usage NCSA/TG Allocations User and project packets Frost (IBM BG/L) TGCDB FEN 1 DBST Usage packets FEN 2 AMIE Users and projects info Service Node FEN 3 ACC 8 Re. SET Job data FEN 4 tgactsync /etc/passwd Users, Projects, DNs /etc/project Re. SET usersync grid-mapfile

High Performance Storage q Redeploying Bluesky storage Special projects and extra storage for the High Performance Storage q Redeploying Bluesky storage Special projects and extra storage for the Tera. Grid Redistributed 5 racks of storage for use with Frost Re-Architected for performance and high availability q The FASt. T storage system - Capacity: 32. 85 TB Raw (~20 TB Usable) Performance: up to 1 GB/s Design and Implementation Full Hardware Redundancy - Full Software Redundancy with Linux multi-path and GPFS -

A FC Switch A B Frost FC Switch B FENs A B A B A FC Switch A B Frost FC Switch B FENs A B A B A B

Continuing Work q MPI-IO testing and performance analysis: GPFS on Frost storage subsystems (FASt. Continuing Work q MPI-IO testing and performance analysis: GPFS on Frost storage subsystems (FASt. T 500 and FASt. T 900) GPFS, Lustre, PVFS 2 on dedicated storage system (Maelstrom) q Semi-Automated Performance Analysis Benchmark Configuration Data Collection Plotting and Analysis q Incorporating these efforts into an upcoming paper on High Performance Storage

Questions? Questions?