statistics/prep.r

   1
   2 source("functions.r");
   3
   4 ikern <- read.csv("/Users/soltesz/Downloads/out.csv", TRUE, sep=",")
   5 f<-factor(ikern$kernel_version, sort(unique(ikern$kernel_version)), sequence(length(unique(ikern$kernel_version))))
   6
   7 u<-ikern$uptime/(60*60*24)
   8
   9 current_time <- as.numeric(format(Sys.time(), "%s"))
  10 i<-(current_time-ikern$install_date)/(60*60*24)
  11
  12 plot(f,u)
  13
  14
  15 sites <- read.csv("/Users/soltesz/Downloads/sites.csv", TRUE, sep=",")
  16 f<-factor(sites$status, sort(unique(sites$status)), sequence(length(unique(sites$status))))
  17
  18 s<-sites$sliver_count
  19
  20 res <- read.csv("/Users/soltesz/Downloads/out_resources.csv", TRUE, sep=",")
  21 library(lattice)
  22 cloud(memsize ~ disksize * cpuspeed|numcores, data=res)
  23
  24 x<-c(res[2],res[4],res[5])
  25 pairs(x)
  26
  27
  28
  29 mdrc <- read.csv("/Users/soltesz/Downloads/out_resources.csv", TRUE, sep=",")
  30
  31 stripchart(round(slices(mdrc)), method="jitter")
  32 hist(round(slices(mdrc)),breaks=30)
  33
  34 hist(round(slices(mdrc)),breaks=30,xlim=c(0,32))
  35 stripchart(round(slices(mdrc)), method="jitter", add=TRUE, jitter=30, at=50)
  36
  37
  38 # bottom, left, top, right
  39 par(mai=c(0,1,0.5,0.2))
  40 hist(round(slices(mdrc)),breaks=30,xlim=c(0,32))
  41 par(mai=c(1.0,1,0.5,0.2))
  42 stripchart(round(slices(mdrc))-0.5, method="jitter", jitter=20, xlim=c(0,32), ylim=c(-25,25),  ylab="Raw Samples", xlab="Slice count as a function of Mem, CPU, Disk")
  43
  44
  45 png("/Users/soltesz/Downloads/slices.png")
  46 par(mfrow=c(2,1))
  47 par(mai=c(0,1,0.5,0.2))
  48 hist(round(slices(mdrc)),breaks=30,xlim=c(0,32), main="Distribution of Slice Count as Function of Mem, CPU, Disk")
  49 par(mai=c(1.0,1,0.5,0.2))
  50 stripchart(round(slices(mdrc))-0.5, method="jitter", jitter=20, xlim=c(0,32), ylim=c(-25,25),  ylab="Raw Samples", xlab="Slice count as a function of Mem, CPU, Disk for live Planetlab Machines")
  51 dev.off()
  52
  53
  54 #-----------------------
  55
  56 f<-slices
  57 f<-slices_2
  58
  59 s2<- f(mdrc, FALSE);
  60 mdrc$score <- s2;
  61 df <- data.frame(mdrc);
  62 b<-30;
  63
  64 # ----------------------
  65 ### LOGINBASE
  66 unique_loginbase_length <- length(unique(mdrc$loginbase));
  67 unique_lb <- list(loginbase=array(0,c(unique_loginbase_length)),
  68                                   score=array(0,c(unique_loginbase_length)),
  69                                   memsize=array(0,c(unique_loginbase_length)),
  70                                   disksize=array(0,c(unique_loginbase_length)),
  71                                   cpuspeed=array(0,c(unique_loginbase_length))
  72                                   )
  73
  74 for ( i in 1:length(mdrc$loginbase) )
  75 {
  76     r <- mdrc[i,];
  77         v <- f(r, TRUE);
  78         unique_lb$loginbase[r$loginbase] <- r$loginbase;
  79         unique_lb$score[r$loginbase]    <- unique_lb$score[r$loginbase]  + r$score;
  80 }
  81
  82 for ( i in 1:length(mdrc$loginbase) )
  83 {
  84     r <- mdrc[i,];
  85         v <- f(r, TRUE);
  86         rscore <- unique_lb$score[r$loginbase]
  87         unique_lb$memsize[r$loginbase]  <- unique_lb$memsize[r$loginbase]  + v[1];
  88         unique_lb$disksize[r$loginbase] <- unique_lb$disksize[r$loginbase]  + v[2];
  89         unique_lb$cpuspeed[r$loginbase] <- unique_lb$cpuspeed[r$loginbase]  + v[3];
  90 }
  91
  92 df<- data.frame(unique_lb)
  93
  94 h<- hist(df$score, breaks=b);
  95 bins<-max(length(h$breaks),max(h$breaks));
  96 c<- array(0,c(bins));
  97 d<- array(0,c(bins));
  98 m<- array(0,c(bins));
  99 # foreach score value, find which range it falls into,
 100 # then in three columns for cpu, mem, disk, record the fraction of each.
 101 # then plot each sequence in a stacked graph, perhaps beside h$counts
 102 for ( i in 1:length(df$cpuspeed) )
 103 {
 104     r <- df[i,];
 105     s <- index_of_bin(h, r$score); # find bin position...
 106     # take fraction that each component contributes to the total, and add to sum
 107
 108     m[s] <- m[s] + unique_lb$memsize[r$loginbase];
 109     d[s] <- d[s] + unique_lb$disksize[r$loginbase];
 110     c[s] <- c[s] + unique_lb$cpuspeed[r$loginbase];
 111 }
 112
 113 # ----------------------
 114 ### HOSTS
 115 # ---  get plot of contributing parts
 116 h<- hist(df$score, breaks=b);
 117 bins<-max(length(h$breaks),max(h$breaks));
 118 c<- array(0,c(bins));
 119 d<- array(0,c(bins));
 120 m<- array(0,c(bins));
 121 # foreach score value, find which range it falls into,
 122 # then in three columns for cpu, mem, disk, record the fraction of each.
 123 # then plot each sequence in a stacked graph, perhaps beside h$counts
 124 for ( i in 1:length(df$cpuspeed) )
 125 {
 126     r <- df[i,1:6];
 127     s <- index_of_bin(h, r$score); # find bin position...
 128     # take fraction that each component contributes to the total, and add to sum
 129     v <- f(r, TRUE);
 130     m[s] <- m[s] + v[1]/r$score;
 131     d[s] <- d[s] + v[2]/r$score;
 132     c[s] <- c[s] + v[3]/r$score;
 133 }
 134
 135
 136 #a <- array(c(c,d,m), dim=c(bins, 3));
 137 a <- array(c(c), dim=c(bins, 3));
 138
 139 #png("/Users/soltesz/Downloads/slice_policy_1.png")
 140 par(mfrow=c(2,1))
 141 par(mai=c(0.5,1,0.5,0.2))
 142 barplot(c(0,h$counts),
 143     xlab="slice count",
 144     main="Distribution of Per-node 'Scores' Calculated from Mem/Disk/CPU",
 145     ylab="Total Frequency",
 146     ylim=c(0,160))
 147 par(mai=c(1.0,1,0,0.2));
 148 barplot(t(a),
 149     legend=c("CPUspeed (GHz)", "DISKsize (GB)", "MEMsize (GB)"),
 150     col=c("pink", "lightblue", "lightgreen"),
 151     ylim=c(0,160),
 152     ylab="Total with Break-down",
 153     xlab="Per-node Score",
 154     names.arg=h$breaks,
 155 );
 156 #dev.off()
 157
 158
 159
 160 #a <- list(cpuspeed=c, memsize=m, disksize=d);
 161 # barplot(t(a), legend=c("cpuspeed", "disksize", "memsize"), col = c("pink", "lightblue", "lightgreen"), ylab="Total Contribution by CPU, Disk, Mem ")