In [6]:

    

using Gadfly; using DataFrames; using GLM;


    

In [3]:

    

#=
tests done on 2 nodes; 
1 local i7 2.6 GHz quadcore with hyperthreading running windows and the registry
1 remote 2Ghz i5 dualcore with hyperthreading running linux

Notable is the linux machine burning more computations than the local windows machine. 
=#
data = readtable("results-03-01-3506.csv", eltypes=[Int64,Int64,Int64,Int64,Int64,Int64]);


    

In [4]:

    

data_means = groupby(data, [:n,:split,:line_consumers,:line_searchers, :search_segments]); 
data_means = combine(map(d -> median(d[:time_taken]) / 1000000, data_means))
rename!(data_means, :x1, :ms_time )
sort!(data_means, cols=[:ms_time])


    

    
Out[4]:




n
split
line_consumers
line_searchers
search_segments
ms_time
1
10000
8000
1
3
4
72.5024515
2
10000
8000
2
2
4
84.9465735
3
10000
2000
2
2
4
120.755968
4
10000
2000
1
3
4
159.355899
5
10000
500
2
2
4
165.8729415
6
10000
500
1
3
4
186.5051135
7
40000
8000
2
2
4
366.142048
8
40000
2000
2
2
4
421.4802675
9
40000
8000
1
3
4
441.293911
10
40000
2000
1
3
4
476.553408
11
40000
500
2
2
4
477.0026415
12
40000
500
1
3
4
536.3740945
13
160000
8000
2
2
4
1594.253562
14
160000
2000
2
2
4
1610.5564335
15
160000
8000
1
3
4
1635.9254915
16
160000
2000
1
3
4
1667.8609385
17
160000
500
2
2
4
1779.6316445
18
160000
500
1
3
4
2001.7535105

In [9]:

    

plot(data_means[(data_means[:line_searchers].==2), :], x=:split, y=:ms_time, color=:n, Geom.point, Geom.line, Scale.x_continuous(format=:plain),Scale.y_continuous(format=:plain))


    

    
Out[9]:

In [100]:

    

plot(data_means[(data_means[:split].==8000), :], x=:line_searchers, y=:ms_time, color=:n, Geom.point, Geom.line, Scale.x_continuous(format=:plain),Scale.y_continuous(format=:plain))


    

    
Out[100]:

In [10]:

    

plot(data_means[(data_means[:n].==160000), :], x=:line_consumers, y=:ms_time, color=:split, Geom.point, Geom.line, Scale.x_continuous(format=:plain),Scale.y_continuous(format=:plain))


    

    
Out[10]:

In [112]:

    

using GLM


    

In [118]:

    

OLS = glm(@formula(time_taken ~ n + split + line_searchers), data, Normal(), IdentityLink());
predict(OLS)


    

    
Out[118]:





360-element Array{Float64,1}:
 1.69777e8
 1.69777e8
 1.69777e8
 1.69777e8
 1.69777e8
 1.69777e8
 1.69777e8
 1.69777e8
 1.69777e8
 1.69777e8
 1.69777e8
 1.69777e8
 1.69777e8
 ⋮        
 1.66667e9
 1.66667e9
 1.66667e9
 1.66667e9
 1.66667e9
 1.66667e9
 1.66667e9
 1.66667e9
 1.66667e9
 1.66667e9
 1.66667e9
 1.66667e9