-
Notifications
You must be signed in to change notification settings - Fork 20
/
Copy pathTrend_Line_Analysis.R
296 lines (229 loc) · 12.3 KB
/
Trend_Line_Analysis.R
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
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
# Exascalar Data Trend Plot
# This program pulls in all the data and then plots a trend of the top performance, median exascalar and top green systems.
## This program imports cleaned data from the Green500 and Top500 lists
## GET THE CLEANED DATA
##check for Exascalar Directory. If none exists stop program with error
##check to ensure results director exists
if(!file.exists("./results")) stop("Data not found in directory Exascalar, first run Exascalar_Cleaner to get tidy data")
d=getwd()
## set working directory
# define Data Directories to use
results <- "./results"
## ------------------------
ExaPerf <- 10^12 ##in Megaflops
ExaEff <- 10^12/(20*10^6) ##in Megaflops/Watt
## this function coputes Exascalar from a list with columns labeled $rmax and $megaflopswatt
## note the function computes to three digits explicitly
compute_exascalar <- function(xlist){
## compute exascalar
t1 <- (log10(xlist$rmax*10^3/ExaPerf) + log10(xlist$mflopswatt/(ExaEff)))/sqrt(2)
## round to three digits
t2 <- round(t1, 3)
## clean up
format(t2, nsmall=3)
}
## Read results files
# import data set
## there are probably ways to simplify this code but this brute force method is easy to read.
Nov14 <- read.csv(paste0(results, "/Nov14.csv"), header=TRUE)
Jun14 <- read.csv(paste0(results, "/Jun14.csv"), header=TRUE)
Nov13 <- read.csv(paste0(results, "/Nov13.csv"), header=TRUE)
Jun13 <- read.csv(paste0(results, "/Jun13.csv"), header=TRUE)
Nov12 <- read.csv(paste0(results, "/Nov12.csv"), header=TRUE)
Jun12 <- read.csv(paste0(results, "/Jun12.csv"), header=TRUE)
Nov11 <- read.csv(paste0(results, "/Nov11.csv"), header=TRUE)
Jun11 <- read.csv(paste0(results, "/Jun11.csv"), header=TRUE)
Nov10 <- read.csv(paste0(results, "/Nov10.csv"), header=TRUE)
Jun10 <- read.csv(paste0(results, "/Jun10.csv"), header=TRUE)
Nov09 <- read.csv(paste0(results, "/Nov09.csv"), header=TRUE)
Jun09 <- read.csv(paste0(results, "/Jun09.csv"), header=TRUE)
#Nov08 <- read.csv(paste0(results, "/green500_top_200811.csv"), header=TRUE)
#Jun08 <- read.csv(paste0(results, "/green500_top_200806.csv"), header=TRUE)
print("data read")
##PLOT MID, MEDIAN AND TOP EXASCALAR TREND
TopEx <- rbind(Jun09[1,1:10], Nov09[1,1:10], Jun10[1,1:10], Nov10[1,1:10], Jun11[1,1:10], Nov11[1,1:10], Jun12[1,1:10], Nov12[1,1:10], Jun13[1,1:10],
Nov13[1,1:10], Jun14[1,1:10], Nov14[1,1:10])
TopGreen500<-rbind(Jun09[Jun09$green500rank==1,1:10], Nov09[Nov09$green500rank==1,1:10],
Jun10[Jun10$green500rank==1,1:10], Nov10[Nov10$green500rank==1,1:10],
Jun11[Jun11$green500rank==1,1:10], Nov11[Nov11$green500rank==1,1:10],
Jun12[Jun12$green500rank==1,1:10], Nov12[Nov12$green500rank==1,1:10],
Jun13[Jun13$green500rank==1,1:10], Nov13[Nov13$green500rank==1,1:10],
Jun14[Jun14$green500rank==1,1:10], Nov14[Nov14$green500rank==1,1:10])
TopGreen500<-TopGreen500[complete.cases(TopGreen500),]
TopGreen500<-TopGreen500[!duplicated(TopGreen500$date),] #get rid of duplicate cases
TopTop500<-rbind(Jun09[Jun09$top500rank==1,1:10], Nov09[Nov09$top500rank==1,1:10],
Jun10[Jun10$top500rank==1,1:10], Nov10[Nov10$top500rank==1,1:10],
Jun11[Jun11$top500rank==1,1:10], Nov11[Nov11$top500rank==1,1:10],
Jun12[Jun12$top500rank==1,1:10], Nov12[Nov12$top500rank==1,1:10],
Jun13[Jun13$top500rank==1,1:10], Nov13[Nov13$top500rank==1,1:10],
Jun14[Jun14$top500rank==1,1:10], Nov14[Nov14$top500rank==1,1:10])
TopTop500<-TopTop500[complete.cases(TopTop500),]
TopTop500<-TopTop500[!duplicated(TopTop500$date),] #get rid of duplicate cases
LowestPower<-rbind(Jun09[Jun09$power==min(Jun09$power),1:10], Nov09[Nov09$power==min(Nov09$power),1:10],
Jun10[Jun10$power==min(Jun10$power),1:10], Nov10[Nov10$power==min(Nov10$power),1:10],
Jun11[Jun11$power==min(Jun11$power),1:10], Nov11[Nov11$power==min(Nov11$power),1:10],
Jun12[Jun12$power==min(Jun12$power),1:10], Nov12[Nov12$power==min(Nov12$power),1:10],
Jun13[Jun13$power==min(Jun13$power),1:10], Nov13[Nov13$power==min(Nov13$power),1:10],
Jun14[Jun14$power==min(Jun14$power),1:10], Nov14[Nov14$power==min(Nov14$power),1:10])
MaximumPower<-rbind(Jun09[Jun09$power==max(Jun09$power),1:10], Nov09[Nov09$power==max(Nov09$power),1:10],
Jun10[Jun10$power==max(Jun10$power),1:10], Nov10[Nov10$power==max(Nov10$power),1:10],
Jun11[Jun11$power==max(Jun11$power),1:10], Nov11[Nov11$power==max(Nov11$power),1:10],
Jun12[Jun12$power==max(Jun12$power),1:10], Nov12[Nov12$power==max(Nov12$power),1:10],
Jun13[Jun13$power==max(Jun13$power),1:10], Nov13[Nov13$power==max(Nov13$power),1:10],
Jun14[Jun14$power==max(Jun14$power),1:10], Nov14[Nov14$power==max(Nov14$power),1:10])
##mean efficiency function calculated the mean perforance adn power and then the mean efficiency from that ratio
##thus defined it reflects the popultion of the Top500 computers.
mean_eff <- function(list){mean(list$rmax)/mean(list$power)}
MeanEx <- matrix(c(mean_eff(Jun09), mean_eff(Nov09),
mean_eff(Jun10), mean_eff(Nov10),
mean_eff(Jun11), mean_eff(Nov11),
mean_eff(Jun12), mean_eff(Nov12),
mean_eff(Jun13), mean_eff(Nov13),
mean_eff(Jun14), mean_eff(Nov14),
mean(Jun09$rmax), mean(Nov09$rmax),
mean(Jun10$rmax), mean(Nov10$rmax),
mean(Jun11$rmax), mean(Nov11$rmax),
mean(Jun12$rmax), mean(Nov12$rmax),
mean(Jun13$rmax), mean(Nov13$rmax),
mean(Jun14$rmax), mean(Nov14$rmax)),
ncol=2, nrow = 12)
MeanEx <- as.data.frame(MeanEx)
names(MeanEx) <- c("mflopswatt", "rmax")
median_eff <- function(list){median(list$rmax)/median(list$power)}
MedianEx <- matrix(c(median_eff(Jun09), median_eff(Nov09),
median_eff(Jun10), median_eff(Nov10),
median_eff(Jun11), median_eff(Nov11),
median_eff(Jun12), median_eff(Nov12),
median_eff(Jun13), median_eff(Nov13),
median_eff(Jun14), median_eff(Nov14),
median(Jun09$rmax), median(Nov09$rmax),
median(Jun10$rmax), median(Nov10$rmax),
median(Jun11$rmax), median(Nov11$rmax),
median(Jun12$rmax), median(Nov12$rmax),
median(Jun13$rmax), median(Nov13$rmax),
median(Jun14$rmax), median(Nov14$rmax)),
ncol=2, nrow = 12)
MedianEx <- as.data.frame(MedianEx)
names(MedianEx) <- c("mflopswatt", "rmax")
bottom_eff <- function(list){list$rmax[which(list$X == max(list$X))]/list$power[which(list$X == max(list$X))]}
bottom_perf <- function(list){list$rmax[which(list$X == max(list$X))]}
BottomGreen <- matrix(c(bottom_eff(Jun09), bottom_eff(Nov09),
bottom_eff(Jun10), bottom_eff(Nov10),
bottom_eff(Jun11), bottom_eff(Nov11),
bottom_eff(Jun12), bottom_eff(Nov12),
bottom_eff(Jun13), bottom_eff(Nov13),
bottom_eff(Jun14), bottom_eff(Nov14),
bottom_perf(Jun09), bottom_perf(Nov09),
bottom_perf(Jun10), bottom_perf(Nov10),
bottom_perf(Jun11), bottom_perf(Nov11),
bottom_perf(Jun12), bottom_perf(Nov12),
bottom_perf(Jun13), bottom_perf(Nov13),
bottom_perf(Jun14), bottom_perf(Nov14)),
ncol=2, nrow = 12)
BottomGreen <- as.data.frame(MedianEx)
names(BottomGreen) <- c("mflopswatt", "rmax")
top_eff <- function(list){list$rmax[which(list$green500rank == 1)[1]]/list$power[which(list$green500rank == 1)[1]]}
top_perf <- function(list){list$rmax[which(list$green500rank == 1)[1]]}
TopGreen <- matrix(c(top_eff(Jun09), top_eff(Nov09),
top_eff(Jun10), top_eff(Nov10),
top_eff(Jun11), top_eff(Nov11),
top_eff(Jun12), top_eff(Nov12),
top_eff(Jun13), top_eff(Nov13),
top_eff(Jun14), top_eff(Nov14),
top_perf(Jun09), top_perf(Nov09),
top_perf(Jun10), top_perf(Nov10),
top_perf(Jun11), top_perf(Nov11),
top_perf(Jun12), top_perf(Nov12),
top_perf(Jun13), top_perf(Nov13),
top_perf(Jun14), top_perf(Nov14)),
ncol=2, nrow = 12)
TopGreen <- as.data.frame(TopGreen)
names(TopGreen) <- c("mflopswatt", "rmax")
DatesString<-c("06/01/2009", "11/01/2009","06/01/2010","11/01/2010","06/01/2011","11/01/2011",
"06/01/2012",
"11/01/2012","06/01/2013",
"11/01/2013","06/01/2014", "11/01/2014")
Date <- as.Date(DatesString, "%m/%d/%Y")
#TopExTrend<-as.data.frame(cbind(Date, exascalar))
#TopExTrend$Date <- format(TopExTrend$Date, format = "%B %d %Y")
##PlotMean over Exascalar Data
## EXASCALAR TREND
## Plot of the Top and Median Exascalar for current cleaned data set
require(ggplot2)
## create TopEx vector and fit
topexascalar<-TopEx$exascalar
## create TopEX data frame for fitting
TopExData <- as.data.frame(cbind(Date, topexascalar))
## fitted model of Top Exascalar data
TopExFit <- lm(topexascalar ~ Date , data = TopExData)
##Create Top Green data and fit
topgreenexascalar<-TopGreen500$exascalar
##create Top Green data frame for fitting
TopGreenExData <- as.data.frame(cbind(Date, topgreenexascalar))
TopGreenExFit <- lm(topgreenexascalar ~ Date , data = TopGreenExData)
## create median vector and fit
MedianEx$exascalar <- compute_exascalar(MedianEx)
medianexascalar <- as.numeric(MedianEx$exascalar)
## createe median data fram for fitting
MedianExData <- as.data.frame(cbind(Date, medianexascalar))
##fitted model of median data
MedianExFit <- lm(medianexascalar ~ Date , data = MedianExData)
##PLOT THE DATA
## plot the data
png(filename= "ExascalarTrendFit.png", height=300, width=400)
plot(Date, topexascalar,
ylim=c(-7.0,0),
xlim = c(14000, 19000),
main = "",
ylab = "Exascalar",
col = "red",
bg = "steelblue2",
pch=21)
par(new=TRUE)
#png(filename= "Exascalar_Max_Med_Trend.png", height=300, width=485)
plot(Date, medianexascalar,
ylim=c(-7.0,0),
xlim = c(14000, 19000),
xlab = "",
ylab = "",
col = "dark blue",
bg = "green",
pch=19)
par(new=TRUE)
plot(as.Date(TopGreen500$date), TopGreen500$exascalar,
ylim=c(-7.0,0),
xlim = c(14000, 19000),
xlab = "",
ylab = "",
main = "Nov 2014 Exascalar Trend",
col = "darkgreen",
bg = "green",
pch=20)
## get parameters for fitted lines
topslope<-TopExFit$coefficient[2]
topintercept<-TopExFit$coefficient[1]
topgreenslope<-TopGreenExFit$coefficient[2]
topgreenintercept<-TopGreenExFit$coefficient[1]
## calculate date zero - when the top trend will intercet zero exascalar
## the zero date is an important figure of merit of the population (zero exascalar)
## representing the most advanced supercomputing capability
datezero = -topintercept/topslope
##draw lines
lines(c(14000, datezero), c(topintercept+topslope*14000, topintercept+topslope*datezero))
medianslope<-MedianExFit$coefficient[2]
medianintercept<-MedianExFit$coefficient[1]
greenslope<-TopGreen500ExFit$coefficient[2]
greenintercept<-TopGreen500ExFit$coefficient[1]
## draw fitted line for median
lines(c(14000, datezero), c(medianintercept+medianslope*14000, medianintercept+medianslope*datezero))
lines(c(14000, datezero), c(topgreenintercept+topgreenslope*14000, topgreenintercept+topgreenslope*datezero))
## add text to graph
text(datezero, 0, as.Date(datezero, origin="1970-01-01"), cex=.7, srt=0, pos = 2)
text(datezero, -.5, "Top Performance", cex=.8, srt=20, pos = 2)
text(datezero, medianintercept+medianslope*datezero-.5, "Median", cex=.8, srt=13, pos = 2)
text(datezero, topgreenintercept+topgreenslope*datezero-.5, "Top Efficiency", cex=.8, srt=13, pos = 2)
text(datezero-1500,
-7, "data from Green500 and Top500",
cex=.6, col="black", pos=3)
dev.off()
print("ExascalarTrend done")