[Disksim-users] problem submitting concurrent I/O requests to DiskSim under system simulator (slave) mode

Sagar Thapaliya sagar at uab.edu
Tue Mar 31 17:13:32 EDT 2015


Hi,
I am trying to understand how to use DiskSim for a shared system, while using it under system simulator (i.e. DiskSim as slave). I am new to DiskSim, so had problem figuring something out, and wondering if I'm even doing it the right way. I would really appreciate if you would be able to help me with the following question and problem I faced:


Question on concurrent access:

Is it possible to submit a new request to DiskSim before an outstanding request is processed? Here, I basically want to create a situation where there are multiple concurrent IO request in the disk queue.



Details:

And here are details of what I was trying to achieve, steps I have tried with syssim_driver.c code, and the problem I have faced:


----------

Original arrival of syssim:

I noticed that the example code syssim_driver.c generates new IO requests only after an existing request is  processed till its completion.


Modified arrival with overlap:

I changed the example code to add second requests before the first request is complete (e.g.,  before calling disksim_interface_internal_event).


Used disksim_request.id to track request's progress:

I also wanted an ability to track when a specific I/O request completed. So, I utilized the field "id" of disksim_request: set a unique value to each request in main method, and read values "id" and completion time "t" in method syssim_report_completion.


Problem:

-  Original arrival of syssim: worked fine: id and t were unique for each requests, and values looked correct.

-  Modified arrival with overlap:  Had this problem - each time syssim_report_completion was invoked, its parameter disksim_request *r seemed to have same value (r.id, r.bytecount, r.start).


I performed these tests with disksim 4.0, with hdd and ssd models (FlashSim).

----------


Could you suggest if I am doing anything wrong here? Or is it possible that this problem is coming due to known limitation of Disksim?





Regards,
Sagar Thapaliya
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.andrew.cmu.edu/pipermail/disksim-users/attachments/20150331/ad37e9d9/attachment.html>


More information about the Disksim-users mailing list