The high performance storage system
Proceedings of the 1993 ACM/IEEE conference on Supercomputing
Efficient organization and access of multi-dimensional datasets on tertiary storage systems
Information Systems - Special issue: scientific databases
Server-directed collective I/O in Panda
Supercomputing '95 Proceedings of the 1995 ACM/IEEE conference on Supercomputing
Network attached storage architecture
Communications of the ACM
Managing energy and server resources in hosting centers
SOSP '01 Proceedings of the eighteenth ACM symposium on Operating systems principles
Energy Management for Server Clusters
HOTOS '01 Proceedings of the Eighth Workshop on Hot Topics in Operating Systems
ICCS'05 Proceedings of the 5th international conference on Computational Science - Volume Part II
| Hi-index | 0.00 |
Tera-scale high-performance computing has enabled scientists to tackle very large and computationally challenging scientific problems, making the advancement of scientific discovery at a faster pace. However, as computing scales to levels never seen before, it also becomes extremely data intensive, I/O intensive, and energy consuming. Amongst these, I/O is becoming a major bottleneck, impeding the expected pace of scientific discovery and analysis of data. Furthermore, the applications are becoming increasingly dynamic in terms of their computation patterns as well as data access patterns to cope with larger problems and data sizes. Due to the complexities of systems and applications and their high energy consumptions, it is, therefore, very important to address research issues and develop dynamic techniques at the level of run-time systems and compilers to scale I/O in the right proportions. This paper presents the details of a dynamic compilation framework developed specifically for I/O-intensive large-scale applications. Our dynamic compilation framework includes a set of powerful I/O optimizations designed to minimize execution cycles and energy consumption, and generates results that are competitive with hand-optimized codes in terms of energy consumption.