[Disksim-users] Problem about stopping simulation because of saturation

[Peter Macko]

Hi,

This is most likely because of DiskSim's hardcoded limit on the number of requests that can be in flight at the same time. The maximum number is MAX_QUEUE_LENGTH declared in src/disksim_logorg.c, which is by default 10000.

The problem is that disksim replays the disk I/Os from the trace at the same speed regardless of the performance of the simulated device. For example, if the trace has an I/O every 1ms, then disksim will issue an I/O every 1ms, regardless of how long that I/O takes to complete - even if it is not yet complete. So if the trace issues requests faster than the device can process, the simulation will die because of saturation.

A quick solution for this problem would be to increase the simulation time between successive requests. Multiplying all timestamps by 2 or 3 should help.

A more general solution to the problem - but also by far the most time solution - would be to develop a new trace format, with the corresponding record and replay tools, in which disksim would know which I/Os need to complete before which I/Os can be issued, so that the timing of the requests in the trace would adjust dynamically depending on the device performance. I hoped to do it at some point, but I was just never able to find the time for it.

Best,

-Peter


On Apr 12, 2012, at 5:35 AM, cq25062030 wrote:

> Hi all,
> 
> I'm a freshman  of using DiskSim. Recently, I try to implement a simply Garbage Collection algorithm based on the ssd model source code provide by Microsoft, I only modify a part of  the ssd_clean.c file, and make a little change about the initialization function. The source code of DiskSim is kept the same as before. And I use the same test case as  Microsoft provided, but when I try to run the new code, the DiskSim give an error message like: Stopping simulation because of saturation..... 
> 
> This problem occurs when run the ssd-rw5m and ssd-postmark  test case.
> 
> I try to reduce the size of the request by modifying the configuration file as the website
> 
> https://sos.ece.cmu.edu/pipermail/disksim-users/2010-July/000516.html, 
> 
> the number of  simulated  request grows, but the problem occurs again at last. Also I try to debug th e program, I find that the requests keep arriving, but little operation about issuing them, I can't find the reason that leads to generate this problem.
> 
> Could anyone give me some advice to solve  this problem? 
> 
> Thanks a lot.
> 
> Best wishes,
> 
> Qian Cheng.
> 
> 
> 
> 
> 
> _______________________________________________
> Disksim-users mailing list
> Disksim-users at ece.cmu.edu
> https://sos.ece.cmu.edu/mailman/listinfo/disksim-users

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.andrew.cmu.edu/pipermail/disksim-users/attachments/20120412/62c87051/attachment.html>