statistics/site_scores_bw.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_3
  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                                   bwlimit=array(0,c(unique_loginbase_length))
  27                           )
  28
  29 for ( i in 1:length(mdrc$loginbase) )
  30 {
  31     r <- mdrc[i,];
  32         unique_lb$loginbase[r$loginbase] <- r$loginbase;
  33         unique_lb$score[r$loginbase]     <- unique_lb$score[r$loginbase] + r$score;
  34
  35         v <- f(r, TRUE);
  36         unique_lb$memsize[r$loginbase]  <- unique_lb$memsize[r$loginbase]  + v[1];
  37         unique_lb$disksize[r$loginbase] <- unique_lb$disksize[r$loginbase]  + v[2];
  38         unique_lb$cpuspeed[r$loginbase] <- unique_lb$cpuspeed[r$loginbase]  + v[3];
  39         unique_lb$bwlimit[r$loginbase] <- unique_lb$bwlimit[r$loginbase]  + v[4];
  40 }
  41
  42 df<- data.frame(unique_lb)
  43
  44 h<- hist(df$score, breaks=b);
  45 bins<-length(h$breaks);
  46 c<- array(0,c(bins));
  47 d<- array(0,c(bins));
  48 m<- array(0,c(bins));
  49 b<- array(0,c(bins));
  50 # foreach score value, find which range it falls into,
  51 # then in three columns for cpu, mem, disk, record the fraction of each.
  52 # then plot each sequence in a stacked graph, perhaps beside h$counts
  53 for ( i in 1:length(df$cpuspeed) )
  54 {
  55     r <- df[i,];
  56     s <- index_of_bin(h, r$score); # find bin position...
  57     # take fraction that each component contributes to the total, and add to sum
  58
  59     m[s] <- m[s] + unique_lb$memsize[r$loginbase]/r$score;
  60     d[s] <- d[s] + unique_lb$disksize[r$loginbase]/r$score;
  61     c[s] <- c[s] + unique_lb$cpuspeed[r$loginbase]/r$score;
  62     b[s] <- b[s] + unique_lb$bwlimit[r$loginbase]/r$score;
  63 }
  64
  65 #vals <- list(bwlimit=b,cpuspeed=c,disksize=d,memsize=m)
  66 a <- array(c(b,c,d,m), dim=c(bins, 4));
  67
  68 png("/Users/soltesz/Downloads/slice_policy_4.png")
  69 par(mfrow=c(2,1))
  70 par(mai=c(0.5,1,0.5,0.2))
  71 barplot(c(0,h$counts),
  72     xlab="slice count",
  73     main="Distribution of Site Scores",
  74     ylab="Total Frequency",
  75     ylim=c(0,70))
  76 par(mai=c(1.0,1,0,0.2));
  77 barplot(t(a),
  78     legend=c("BWlimit (Mbps)", "CPUspeed (GHz)", "DISKsize (GB)", "MEMsize (GB)"),
  79     col=c("lightyellow", "pink", "lightblue", "lightgreen"),
  80     ylim=c(0,70),
  81     ylab="Break-down by Resource",
  82     xlab="Site Score",
  83     names.arg=c(0,h$breaks[1:length(h$breaks)-1]),
  84 );
  85 dev.off()
  86
  87