glb_scripts/geo_PMR_b.m at master · geographicalloadbalancing/glb_scripts · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
% load data input
has_quadprog = exist( 'quadprog' );
has_quadprog = has_quadprog == 2 | has_quadprog == 3;
has_linprog  = exist( 'linprog' );
has_linprog  = has_linprog == 2 | has_linprog == 3;
rnstate = randn( 'state' ); randn( 'state', 1 );
s_quiet = cvx_quiet(true);
s_pause = cvx_pause(false);
cvx_solver sdpt3;
cvx_clear;
clc;
clear;

a = 10;
b = 48;
I = ones(a,1);
J = ones(b,1);


% workload traces
length = 288;
scale = 1000;
L0 = load('traces/sapTrace.tab');
L = L0(:,4)/scale;
for i = 1:1:length*2
    x1(i) = i/6;
    load1(i) = (L(2*i-1) + L(2*i)) / 2;
end
load1 = load1.^2;
max(load1)/mean(load1)
geography = [930 1536 585 8614 1354 931 213 4921 2261 330 3137 1498 767 700 960 875 352 1443 1710 2334 1407 523 1344 241 455 654 399 2252 441 4596 2015 162 2702 762 1031 3038 278 976 197 1333 5040 623 181 2049 1091 369 1556 132];
time_zone = [2 1 2 0 1 3 3 3 3 1 2 3 2 2 3 2 3 3 3 3 2 2 2 1 2 0 3 3 1 3 3 2 3 2 0 3 3 3 2 2 2 1 3 3 0 3 2 1];
for i=1:1:48
    load2(i,:) = geography(i)*load1(time_zone(i)*6+1:time_zone(i)*6+length);
end


% propagation delay
datacenter_location = [37 120; 47 120; 44 120; 40 90; 31 83; 38 78; 31 99; 28 81; 35 79; 33 81];
state_location = [32 87; 34 119; 35 92; 37 120; 39 105; 41 73; 39 76; 28 81; 31 83; 44 114; 40 89; 40 86; 41 93; 38 98; 38 85; 31 92; 45 69; 39 77; 42 72; 43 84; 45 93; 33 90; 38 92; 47 110; 41 99; 39 116; 43 71; 40 74; 34 106; 43 76; 35 79; 47 100; 40 83; 35 97; 44 120; 41 78; 42 71; 32 80; 44 100; 35 86; 31 99; 40 112; 44 73; 38 79; 47 121; 39 81; 44 89; 43 107];
for i = 1:1:10
    for j = 1:1:48
        delay(i,j)=sqrt((datacenter_location(i,1)-state_location(j,1))^2+(datacenter_location(i,2)-state_location(j,2))^2);
    end
end


% renewable supplies
W1 = load('traces/wind_supply_week.csv');
S1 = load('traces/solar_supply_week.csv');
for i = 1:1:length
    S(i,:) = max(0,S1(2*i-1,:));
    W(i,:) = W1(i,:);
end
s_mean = mean(S(1:length,:));
w_mean = mean(W(1:length,:));
for i = 1:1:a
    S(:,i) = S(:,i)/s_mean(i);
    W(:,i) = W(:,i)/w_mean(i);
end
S = S';
W = W';

% LOCAL
[Y DCM] = min(delay);
DCL = zeros(a,length);
delay_DC = zeros(a,length);
for jo = 1:1:b
    DCL(DCM(jo),:) = DCL(DCM(jo),:) + load2(jo,:);
    delay_DC(DCM(jo),:) = delay_DC(DCM(jo),:) +  load2(jo,:)*delay(DCM(jo),jo);
end
prop_delay_loc = sum(delay_DC')./sum(DCL');


for i = 1:1:1
    i
    beta0 = 6;
    factor = diag(1+1./sqrt(beta0*[10.41 3.73 5.87 7.48 5.86 6.67 6.44 8.6 6.03 5.49]'));
    DCL = factor*DCL;
    M = ceil(2 * max(DCL'))';
    capacity = i * 2 * mean(DCL');
    DCL_total = ones(1,10)*DCL;
    %p = 0;
    for j = 1:1:21
        j
        %wi = (0.75+0.01*(j-1))*i;
        wi = (0.76+0.02*(j-1))*i;
        so = i - wi;
        %re(j) = 0.5*(j-1);
        Re = 2*(S.*(so*capacity'*ones(1,length)) + W.*(wi*capacity'*ones(1,length)));
        for k = 1:6:length
            T2(:,k:k+5) = Re(:,k)*ones(1,6);
        end
        x0 = 0;
        lambda_t = load2';
        mu = ones(a,1);
        energy_cost = [10.41 3.73 5.87 7.48 5.86 6.67 6.44 8.6 6.03 5.49]';
        delay_cost = ones(a,1);
        beta = beta0*ones(a,1);
        prop_delay = delay;
        caps = M;
        w = 3;

        [x_opt(:,:,j,i) cost_opt(j,i) delay_opt(j,i)] = hetero_opt(x0, lambda_t(1:length,:), mu, energy_cost, delay_cost, beta, prop_delay, caps, Re(:,1:length)')
        brown_opt(j,i) = sum(sum(max(0,x_opt(:,:,j,i)-Re(:,1:length)')))


        %for k = 1:1:a
        %    [x_loc(:,k,j,i) cost_local(k) delay_local(k)] = hetero_opt(x0, DCL(k,1:length)', mu(k), energy_cost(k), delay_cost(k), beta(k), prop_delay_loc(1,k), caps(k), Re(k,1:length)');
        %end
        %brown_loc(j,i) = sum(sum(max(0,x_loc(:,:,j,i)-Re(:,1:length)')))
        %delay_loc(j,i) = sum(DCL(:,1:length)')*delay_local(:)/sum(sum(DCL(:,1:length)))
        %cost_loc(j,i) = sum(cost_local(:))

        %csvwrite('results/PMR-compare.csv',[cost_opt, brown_opt, delay_opt],0,0);
        %csvwrite('results/PMR-compare-b.csv',[cost_opt, brown_opt, delay_opt, cost_loc, brown_loc, delay_loc],0,0);
    end
end

%{
figure;
plot(0:0.1:1,brown_opt(1:1:11),'k',0:0.1:1,brown_loc(1:1:11),'b')
xlabel('% of wind');
ylabel('brown energy consumption');
%ylim([0,1]);
legend('GLB','LOCAL');
set (gcf, 'PaperUnits', 'inches', 'PaperPosition', [0.1 0 3.6 2.8]);
print ('-depsc', 'figs/brownWind2.eps');
%}