<div dir="ltr"><b id="internal-source-marker_0.6647921551484615" style="color:rgb(0,0,0);font-family:'Times New Roman';font-size:medium;font-weight:normal"><p dir="ltr" style="margin-top:0pt;margin-bottom:0pt"><span style="font-size:15px;font-family:Arial;background-color:transparent;vertical-align:baseline;white-space:pre-wrap">Hi DiskSim Users,</span></p>

<br><span style="font-size:15px;font-family:Arial;background-color:transparent;vertical-align:baseline;white-space:pre-wrap"></span><p dir="ltr" style="margin-top:0pt;margin-bottom:0pt"><span style="font-size:15px;font-family:Arial;background-color:transparent;vertical-align:baseline;white-space:pre-wrap">I am trying to simulate in-disk scheduling algorithm to determine the effect of interleaving OS IO scheduling and in-disk IO scheduling (scheduling within the disk). To test this feature I modified the scheduling policy bit in diskspecs file and ran the simulation using ascii trace file. However, I noticed that the “Overall I/O System Response time average” along with other parameters remained unaffected from the choice of scheduling algorithm used in disk. These parameters did change when I changed the scheduling policy at the driver level, i.e, in the parv file. </span></p>

<br><span style="font-size:15px;font-family:Arial;background-color:transparent;vertical-align:baseline;white-space:pre-wrap"></span><p dir="ltr" style="margin-top:0pt;margin-bottom:0pt"><span style="font-size:15px;font-family:Arial;background-color:transparent;font-weight:bold;vertical-align:baseline;white-space:pre-wrap">Further Investigation:</span></p>

<br><span style="font-size:15px;font-family:Arial;background-color:transparent;vertical-align:baseline;white-space:pre-wrap"></span><p dir="ltr" style="margin-top:0pt;margin-bottom:0pt"><span style="font-size:15px;font-family:Arial;background-color:transparent;vertical-align:baseline;white-space:pre-wrap">I investigated further and found that “Max queue length” parameter for the disk was set to 1 in diskspecs file. My intuition was that for any scheduling to take place at disk level, it has to queue more than 1 IO request. So I ran the test again after changing this value to x (any where from 2 to 10000).  However, this created another problem. After changing Max queue length,I found that disksim simulated only  36 IO requests (seen from “Overall I/O System Total Requests handled” trace output) out of nearly 10K IO requests in trace file. Note that disksim did not error out while running this test, it simply processed less number of IOs. Also the overall response time did not change with different choice of scheduling algorithm at disk level.</span></p>

<br><span style="font-size:15px;font-family:Arial;background-color:transparent;vertical-align:baseline;white-space:pre-wrap"></span><p dir="ltr" style="margin-top:0pt;margin-bottom:0pt"><span style="font-size:15px;font-family:Arial;background-color:transparent;vertical-align:baseline;white-space:pre-wrap">I am not sure what is causing this problem. Simply put my intention is to run disk-level scheduling and observe output characteristics. I would be glad to share the parv and trace file if someone is interested in running the test on their end.</span></p>

<br><span style="font-size:15px;font-family:Arial;background-color:transparent;vertical-align:baseline;white-space:pre-wrap"></span><p dir="ltr" style="margin-top:0pt;margin-bottom:0pt"><span style="font-size:15px;font-family:Arial;background-color:transparent;vertical-align:baseline;white-space:pre-wrap">Any help is appreciated.</span></p>

<br><span style="font-size:15px;font-family:Arial;background-color:transparent;vertical-align:baseline;white-space:pre-wrap"></span><p dir="ltr" style="margin-top:0pt;margin-bottom:0pt"><span style="font-size:15px;font-family:Arial;background-color:transparent;vertical-align:baseline;white-space:pre-wrap">Thanks,</span></p>

<span style="font-size:15px;font-family:Arial;background-color:transparent;vertical-align:baseline;white-space:pre-wrap">Prashant Saxena </span></b><br></div>