Aspects of Path Sharing in I/OUniversity of California, Santa Cruz, Computer Research Laboratory, 1987 - Computer input-output equipment - 36 pages |
From inside the book
Results 1-3 of 13
Page
... overlapped organization of read transfers . The results obtained show that , as the I / O throughput increases , it is possible for the performance of overlapped I / O to become worse than that of sequential accesses . This is due to ...
... overlapped organization of read transfers . The results obtained show that , as the I / O throughput increases , it is possible for the performance of overlapped I / O to become worse than that of sequential accesses . This is due to ...
Page 1
... transfer can follow the end of the DASD to buffer transfer or , assuming a multi - ported buffer , the two transfers can be overlapped . The overlapped transfer reduces the data transfer portion of the I / O , and thus might seem ...
... transfer can follow the end of the DASD to buffer transfer or , assuming a multi - ported buffer , the two transfers can be overlapped . The overlapped transfer reduces the data transfer portion of the I / O , and thus might seem ...
Page 7
... overlapped transfers . We observe that the overlapped transfer organization results in a better I / O time for lower I / O rates . As the I / O rate increases , however , its advantage over the sequential organization decreases , and ...
... overlapped transfers . We observe that the overlapped transfer organization results in a better I / O time for lower I / O rates . As the I / O rate increases , however , its advantage over the sequential organization decreases , and ...
Common terms and phrases
a₁ approach assume asynchronous disk I/O asynchronous I/O b₁ buffer ports buffer transfer bus bandwidth channel and device channel to buffer channel transfer class i actuators classes of requests coefficient of variation computed conditional expected confidence intervals consider control unit corresponding DASD reconnection DASD workload denote different numbers different speeds disks sharing example expected number groups of disks I/O rate I/O service I/O transfer paths iteration loss system missed reconnection delay n₁ number of class number of request numbers of servers overlapped organization overlapped transfer organization Overlapped vs sequential ovlap paths are busy probability of success propose a simple queueing models r₁ Random Number read access relative error relative performance merits request of class request sources resource Rotational Position Sensing RPS miss seek and latency sequential organization server utilization sharing of I/O simulation results solution string paths success probabilities successful reconnection T₁ total I/O U₁ varying number W₁