[Disksim-users] Why the overall average response time, almost 20000, is so big?

[Yinliang Yue]

I have noticed that the average time interval between two I/O requests is
too small.  The first column means the I/O arrive time (ms), you issued 33
I/Os in 8.8ms.  The average time interval between two I/O requests is only
0.27ms.  However, the disk I/O access time (access time = data transfer
time + rotation time + seek time) is far larger than 0.27ms.  There are too
many pending I/O requests in the waiting queue and so the average response
time is Incredibly large.

2012/8/14 leon hsu <liangxiongxu at gmail.com>

> Hi yueyinliang,
> Hopefully, the name is same with your mail: ) Thanks for you reply.
>
> The attaching file is the hard disk configuration.  I only have one hard
> disk.
>
> Could you take a look and check the configuration for me? Thank you so
> much.
>
> The ascii trace looks like this:
>
> 0    0    0    24    0
> 0    0    0    24    1
> 1    0    24    16    0
> 1.5    0    40    16    0
> 2    0    56    16    0
> 2.2    0    72    16    0
> 2.4    0    88    16    0
> 2.6    0    104    16    0
> 2.8    0    120    16    0
> 3    0    136    8    0
> 3.125    0    144    8    0
> 3.25    0    152    8    0
> 3.375    0    160    8    0
> 3.5    0    168    8    0
> 3.625    0    176    8    0
> 3.75    0    184    8    0
> 3.875    0    192    8    0
> 4    0    200    8    0
> 4.333333    0    208    8    0
> 4.666667    0    216    8    0
> 5    0    224    8    0
> 5.333333    0    232    8    0
> 5.666667    0    240    8    0
> 6    0    248    8    0
> 6.333333    0    256    8    0
> 6.666667    0    264    8    0
> 7    0    272    16    0
> 7.5    0    288    16    0
> 8    0    304    16    0
> 8.2    0    320    16    0
> 8.4    0    336    16    0
> 8.6    0    352    16    0
> 8.8    0    368    16    0
>
>
>
> Leon
>
>
>
> On Sun, Aug 12, 2012 at 11:24 PM, Yinliang Yue <yueyinliang at gmail.com>wrote:
>
>> There are so many I/O requests waiting in the queue. I do not know the
>> details of your storage systems, such as how many disks, how these disks
>> are organized. I suggest you check your storage configuration.
>>
>> 2012/8/13 leon hsu <liangxiongxu at gmail.com>
>>
>>>
>>> Hi, everybody,
>>>
>>> I need you help:-)
>>>
>>> When I set the "I/O Trace Time Scale" in disksim_iosim to 1, the
>>> simulator shows a saturation error before.
>>> But when I change "I/O Trace Time Scale" in disksim_iosim to 10, the
>>> error disappeared.
>>>
>>> I think that is because the "request arrival time" of the trace file is
>>> too small, which results that the number of the arriving requests in one
>>> second is more than ones that the simulator could deal with in one second.
>>> So I change the "I/O Trace Time Scale" in disksim_iosim to 10.
>>>
>>> Next, I run it again, but the weired result made me crazy. When I ran
>>> disksim using my ascii trace file, the overall response time, almost 20000,
>>> is too big to believable.
>>>
>>> Do anybody could tell me the reason and how to solve this problem. Thank
>>> you so much. I am so appreciated.
>>>
>>>
>>> I set the "request arrival time" like the following style.
>>>
>>> 0.5
>>> 0.75
>>> 1
>>> 1.2000
>>> 1.4000
>>> 1.5000
>>> 1.7000
>>> 2.0
>>> .....
>>>
>>>
>>> Hopefully, it could give some hints to find out my problem. Thanks.
>>>
>>>
>>> Leon
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>> _______________________________________________
>>> Disksim-users mailing list
>>> Disksim-users at ece.cmu.edu
>>> https://sos.ece.cmu.edu/mailman/listinfo/disksim-users
>>>
>>>
>>
>>
>> --
>>         此致
>> 敬礼
>>
>> 岳银亮
>>
>>
>
>
> --
>
> Cailiang Xu
>
> Tel: 732.789.6683
>
> Email: liangxiongxu at gmail.com
>
> Computer Architecture & System Lab
> <liangxiongxu at gmail.com>
>
>


-- 
        此致
敬礼

岳银亮
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.andrew.cmu.edu/pipermail/disksim-users/attachments/20120814/7b8456c1/attachment.html>