ROSS R. B., "Reactive Scheduling for Parallel I/O Systems", PhD thesis, Dept. of Electrical and Computer Engineering, Clemson University, Clemson, SC, December 2000. Home/Search Document Details and Download
Summary Related Articles Check |
This paper is cited in the following contexts:
Improved Read Performance in a Cost-Effective.. - Zhu, Jiang, Qin.. (2003) (Correct)
....the I O performance of concurrently executing processes in PVFS is implemented in [12] The role of sensitivity of the I O servers and clients is analyzed in [13] which concludes that when a node serves both as an I O client and as a data server, the overall I O performance will be degraded. In [14, 15], a scheduling scheme is introduced so that the service order of different requests on each server is determined by their desired locations in the space of Logical Block Address and disk arm seeking time is reduced accordingly. In [7] we introduced the design and implementation of CEFT PVFS, and ....
R. B. Ross, "Reactive Scheduling for Parallel I/O Systems," Ph.D. dissertation, Clemson University, December 2000.
Design and Analysis of a Network Transfer Layer for Parallel File.. - Carns (2001) (1 citation) (Correct)
....has increased the knowledge base in the eld of parallel I O. There are many new ideas and implementation lessons that can now be used in parallel le system design. 1.4. 1 Scheduling One of the most important research topics explored in the rst generation PVFS design is Reactive Scheduling [24]. Reactive Scheduling is a new approach to making server side scheduling decisions for parallel le systems. The main goal is to dynamically choose appropriate scheduling policies depending on the state of the system. 6 This is in contrast to the traditional approach of trying to optimize a single ....
Robert B. Ross. Reactive Scheduling for Parallel I/O Systems. PhD thesis, Electrical and Computer Engineering Dept., Clemson University, 2000.
Server-side scheduling in cluster parallel I/O systems - Robert Ross And (2001) (1 citation) Self-citation (Ross) (Correct)
....that no single disk optimizations are used on the I O nodes (mapping PVFS file locations to local file ones is simpler in the single disk case) Varying scheduling algorithms for write workloads showed only limited benefit, so this data is not presented here. Interested readers are directed to [ROS 00] for this data. For read tests, before each run a local data file larger than the size of a node s memory was read in its entirety on each I O node to remove all PVFS file data from cache. A separate run of read tests was additionally performed in which we allowed file data to remain in cache ....
....total amount of data that will be accessed and N blks is the number of blocks into which the data is split (per task) Tests were run for N blks values of 16 and 32. Results for both were similar, so only the results from 32 blocks per task are presented here. Interested readers are directed to [ROS 00] for this data. Figure 10 presents the results for 32 blocks per task. This is the first set of tests for which we see a significant difference in uncached application service times between algorithms. This is a clear indicator that we are reducing the amount of work performed by the underlying ....
ROSS R. B., "Reactive Scheduling for Parallel I/O Systems", PhD thesis, Dept. of Electrical and Computer Engineering, Clemson University, Clemson, SC, December 2000.
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