statistics/site_scores.r

   1
   2 source("functions.r");
   3
   4 #system("../nodequery.py --nodelist > ../nodelist.txt")
   5 #system("../comonquery.py --cache --nodelist ../nodelist.txt --select 'resptime>0' --fields='name,cpuspeed,numcores,memsize,disksize,bwlimit' | grep -v null | ./hn2lb.py | ./hn2pcustatus.py > ./out_resources.csv ")
   6
   7 mdrc <- read.csv("out_resources.csv", TRUE, sep=",")
   8
   9 # replace all weird numbers with defaults of 100mbps
  10 mdrc$bwlimit <- replace(mdrc$bwlimit, which(mdrc$bwlimit==0 | mdrc$bwlimit==1), 100000)
  11
  12 f<-slices_2
  13
  14 s2<- f(mdrc, FALSE);
  15 mdrc$score <- s2;
  16 b<-30;
  17
  18 # ----------------------
  19 ### LOGINBASE
  20 unique_loginbase_length <- length(unique(mdrc$loginbase));
  21 unique_lb <- list(loginbase=array(0,c(unique_loginbase_length)),
  22                                   score=array(0,c(unique_loginbase_length)),
  23                                   memsize=array(0,c(unique_loginbase_length)),
  24                                   disksize=array(0,c(unique_loginbase_length)),
  25                                   cpuspeed=array(0,c(unique_loginbase_length))
  26                           )
  27
  28 for ( i in 1:length(mdrc$loginbase) )
  29 {
  30     r <- mdrc[i,];
  31         unique_lb$loginbase[r$loginbase] <- r$loginbase;
  32         unique_lb$score[r$loginbase]     <- unique_lb$score[r$loginbase] + r$score;
  33
  34         v <- f(r, TRUE);
  35         unique_lb$memsize[r$loginbase]  <- unique_lb$memsize[r$loginbase]  + v[1];
  36         unique_lb$disksize[r$loginbase] <- unique_lb$disksize[r$loginbase]  + v[2];
  37         unique_lb$cpuspeed[r$loginbase] <- unique_lb$cpuspeed[r$loginbase]  + v[3];
  38 }
  39
  40 df<- data.frame(unique_lb)
  41
  42 h<- hist(df$score, breaks=b);
  43 bins<-length(h$breaks);
  44 c<- array(0,c(bins));
  45 d<- array(0,c(bins));
  46 m<- array(0,c(bins));
  47 b<- array(0,c(bins));
  48 # foreach score value, find which range it falls into,
  49 # then in three columns for cpu, mem, disk, record the fraction of each.
  50 # then plot each sequence in a stacked graph, perhaps beside h$counts
  51 for ( i in 1:length(df$cpuspeed) )
  52 {
  53     r <- df[i,];
  54     s <- index_of_bin(h, r$score); # find bin position...
  55     # take fraction that each component contributes to the total, and add to sum
  56
  57     m[s] <- m[s] + unique_lb$memsize[r$loginbase]/r$score;
  58     d[s] <- d[s] + unique_lb$disksize[r$loginbase]/r$score;
  59     c[s] <- c[s] + unique_lb$cpuspeed[r$loginbase]/r$score;
  60 }
  61
  62 a <- array(c(c,d,m), dim=c(bins, 3));
  63
  64 png("/Users/soltesz/Downloads/slice_policy_3.png")
  65 par(mfrow=c(2,1))
  66 par(mai=c(0.5,1,0.5,0.2))
  67 barplot(c(0,h$counts),
  68     xlab="slice count",
  69     main="Distribution of Site Scores",
  70     ylab="Total Frequency",
  71     ylim=c(0,70))
  72 par(mai=c(1.0,1,0,0.2));
  73 barplot(t(a),
  74     legend=c("CPUspeed (GHz)", "DISKsize (GB)", "MEMsize (GB)"),
  75     col=c("pink", "lightblue", "lightgreen"),
  76     ylim=c(0,70),
  77     ylab="Break-down by Resource",
  78     xlab="Site Score",
  79     names.arg=c(0,h$breaks[1:length(h$breaks)-1]),
  80 );
  81 dev.off()
  82
  83