Here are contest benchmarks for 2.5.59-mm10 and 2.5.59-mm10 with the 
anticipatory I/O scheduler in the current experimental patch form. Needs 
tuning.

no_load:
Kernel         [runs]   Time    CPU%    Loads   LCPU%   Ratio
2.5.59              3   79      93.7    0.0     0.0     1.00
2.5.59-mm10         3   79      93.7    0.0     0.0     1.00
2.5.59-mm10ais      1   82      90.2    0.0     0.0     1.00
2.5.59-mm8          3   79      93.7    0.0     0.0     1.00
cacherun:
Kernel         [runs]   Time    CPU%    Loads   LCPU%   Ratio
2.5.59              3   76      97.4    0.0     0.0     0.96
2.5.59-mm10         3   76      97.4    0.0     0.0     0.96
2.5.59-mm10ais      1   77      97.4    0.0     0.0     0.94
2.5.59-mm8          3   76      97.4    0.0     0.0     0.96
process_load:
Kernel         [runs]   Time    CPU%    Loads   LCPU%   Ratio
2.5.59              3   92      81.5    29.7    17.4    1.16
2.5.59-mm10         3   187     39.6    194.0   58.8    2.37
2.5.59-mm10ais      1   181     40.9    187.0   57.7    2.21
2.5.59-mm8          3   193     38.9    200.3   60.1    2.44
ctar_load:
Kernel         [runs]   Time    CPU%    Loads   LCPU%   Ratio
2.5.59              3   98      80.6    2.0     5.1     1.24
2.5.59-mm10         3   99      78.8    1.7     4.0     1.25
2.5.59-mm10ais      1   148     53.4    2.0     5.4     1.80
2.5.59-mm8          3   98      79.6    2.0     5.1     1.24
xtar_load:
Kernel         [runs]   Time    CPU%    Loads   LCPU%   Ratio
2.5.59              3   102     74.5    1.0     3.9     1.29
2.5.59-mm10         3   100     76.0    1.0     4.0     1.27
2.5.59-mm10ais      1   152     50.0    2.0     4.6     1.85
2.5.59-mm8          3   101     75.2    1.0     4.0     1.28
io_load:
Kernel         [runs]   Time    CPU%    Loads   LCPU%   Ratio
2.5.59              3   152     50.0    34.1    13.1    1.92
2.5.59-mm10         3   148     51.4    35.2    14.2    1.87
2.5.59-mm10ais      3   323     24.1    131.0   24.8    3.94
2.5.59-mm8          3   155     49.0    35.6    12.9    1.96
io_other:
Kernel         [runs]   Time    CPU%    Loads   LCPU%   Ratio
2.5.59              3   89      84.3    11.2    5.6     1.13
2.5.59-mm10         3   115     66.1    35.0    18.3    1.46
2.5.59-mm10ais      1   134     58.2    59.7    26.9    1.63
2.5.59-mm8          3   115     67.0    33.4    17.4    1.46
read_load:
Kernel         [runs]   Time    CPU%    Loads   LCPU%   Ratio
2.5.59              3   101     77.2    6.5     5.0     1.28
2.5.59-mm10         3   93      81.7    2.8     2.2     1.18
2.5.59-mm10ais      1   116     67.2    8.2     6.0     1.41
2.5.59-mm8          3   93      81.7    2.8     2.2     1.18
list_load:
Kernel         [runs]   Time    CPU%    Loads   LCPU%   Ratio
2.5.59              3   95      80.0    0.0     6.3     1.20
2.5.59-mm10         3   97      79.4    0.0     6.2     1.23
2.5.59-mm10ais      1   99      78.8    1.0     7.1     1.21
2.5.59-mm8          3   97      79.4    0.0     6.2     1.23
mem_load:
Kernel         [runs]   Time    CPU%    Loads   LCPU%   Ratio
2.5.59              3   95      82.1    52.7    2.1     1.20
2.5.59-mm10         3   97      79.4    53.7    2.0     1.23
2.5.59-mm10ais      1   103     76.7    53.0    1.9     1.26
2.5.59-mm8          3   100     78.0    57.7    2.0     1.27
dbench_load:
Kernel         [runs]   Time    CPU%    Loads   LCPU%   Ratio
2.5.59              3   214     36.4    2.3     40.7    2.71
2.5.59-mm10         3   300     25.3    3.3     41.2    3.80
2.5.59-mm10ais      1   579     13.3    8.0     51.7    7.06
2.5.59-mm8          3   345     22.0    4.3     47.5    4.37

I went back and checked them because I didn't believe the results the first 
time. These are a clean set of results, and they correlate with the first 
lot. The anticipatory scheduler needs tuning. Note that even noload takes 
longer, but the cacherun isn't. All the disk I/O based loads take longer.

Con