Ethan Leo Miller. Storage Hierarchy Management for Scientific Computing. PhD thesis, University of California at Berkeley, 1995. Home/Search Document Not in Database Summary Related Articles Check |
This paper is cited in the following contexts:
RAMA: An easy-to-use, high-performance parallel file system - Miller, Katz (1997) (2 citations) Self-citation (Miller) (Correct)
....node. This approach yields two major advantages: good performance across a wide variety of workloads without data placement hints, and scalability from fewer than ten to hundreds of node disk pairs. This paper provides a brief overview of RAMA; a more complete description may be found in Ref. [17]. RAMA, like most file systems, is I O bound, as disk speeds are increasing less rapidly than network and CPU speeds. While physically small disks are not necessary for RAMA, they reduce hardware cost and complexity by allowing disks to be mounted directly on processor boards rather than ....
....in RAMA by dividing the block offset by an additional hash function parameter s. This scheme yields the same hash value, and thus the same mapping from file block to disk line, for s sequential blocks in a single file. The optimal value for s depends on both disk characteristics and the workload [17]. S is set to 4 for the simulations in this paper. The hash functions used for the experiments described in this paper are based on the multiplicativehash routine from the GNU libg library [14] This hash function is not likely to be the optimal hash function, but optimizations in the hash ....
E.L. Miller, Storage Hierarchy Management for Scientific Computing, Ph.D. thesis, University of California at Berkeley, Jan. 1995.
Towards Scalable Benchmarks for Mass Storage Systems - Miller (1996) Self-citation (Miller) (Correct)
....[6] stressing the storage system in a different way. Programs such as out of core matrix decomposition and global climate modeling make excellent benchmarks because their I O access patterns can easily be captured without the need to actually perform the computations called for in the programs [12]. Rather than actually factor a large matrix, the benchmark simply reads and writes the files in the same pattern as the real application . Similarly, the benchmark simulating global climate modeling does not do any actual modeling, but rather follows the same access pattern as the real program. ....
E. L. Miller, Storage Hierarchy Management for Scientific Computing, Ph.D. dissertation, University of California at Berkeley, April 1995. Also available as technical report UCB/CSD 95/872.
Real-Time Scheduling of Tertiary Storage - Lijding (2003) (Correct)
No context found.
Ethan Leo Miller. Storage Hierarchy Management for Scientific Computing. PhD thesis, University of California at Berkeley, 1995.
Online articles have much greater impact More about CiteSeer.IST Add search form to your site Submit documents Feedback
CiteSeer.IST - Copyright Penn State and NEC