ioprio feature behaviour

[Vasily Tarasov]

Hello.
I produced a little test of ioprio feature. Results are basically good 
enough,
but there is something strange on my mind.
The test just runs 8 simple readers with priorities 0, 1, .., 7 in best 
effort class.
Readers read from files and count how much Mbytes per second they can read.
Results are here:
Process # (prio)  Measurement 1  (Mbps)   Measurement 2 (Mbps)    
Measurement 3 (Mbps)
   0                                 6,60                              
           7,37                                 6,19
   1                                 7,87                              
           7,90                                 7,15
   2                                 5,92                               
          4,75                                 4,61
   3                                 3,31                              
           3,34                                 3,4
   4                                 0,95                               
          0,97                                 1,03
   5                                 1,14                              
           1,23                                 1,2
   6                                 0,83                            
             0,96                                 0,83
   7                                 0,41                            
             0,41                                 0,41
( The whole results are at http://www.7ka.mipt.ru/~vass/cfq-tests/tests.pdf)

The questions are:
1) Why process 0 with priority 0 has less bandwidth than process 1 with 
priority 1?
2) The same with processes (priorities) 4, 5?
3) Why there is no _uniform_ dependence between bandwidth and priority?
4) Why sums of bandwidths of processes when priorities are setted and 
when they are not setted (look in pdf) aren't equal?

Thanks, Vasily.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/