statistics/rt_s3_overhead.r

   1 source("functions.r");
   2
   3
   4
   5 median_time_to_resolve_window <- function (t, tg, window)
   6 {
   7     hbreaks<-tg$week_ts
   8
   9     xx<-NULL;
  10     yy<-NULL;
  11     yy_sd_high<-NULL;
  12     yy_sd_low<-NULL;
  13     date_index <- NULL;
  14     q_list <- NULL;
  15
  16     x<-seq(-20,20,0.01)
  17
  18     for ( i in seq(1,length(hbreaks)-window-1) )
  19     {
  20         print (sprintf("round %s of %s", i, length(hbreaks)-window-1))
  21         # get range from t
  22         t_sub <- t[which(t$start > hbreaks[i] & t$start<= hbreaks[i+window]),]
  23         if ( length(t_sub$start) <= 1 )  { next }
  24         # take log, then sn.mle -> h
  25         d <- (t_sub$lastreply - t_sub$start)/(60*60)    # hours
  26         d <- log(d)                                     # log(hours)
  27             # sn.mle
  28         print (sprintf("length: %s", length(d)))
  29         q<-quantile(d)
  30         print(q)
  31
  32         date_index <- c(date_index, round(i+window/2))
  33
  34         xx<- c(xx, hbreaks[round(i+window/2)])
  35         q_list <- rbind(q_list, q)
  36
  37     }
  38     return (cbind(xx,q_list))
  39 }
  40
  41 available_nodes <- function (ns, from, to, type, fmt="%b")
  42 {
  43     # find 'type' range of days
  44     dates <-seq(as.Date(from), as.Date(to), type)
  45     months <- format(dates, fmt)
  46     hbreaks<-unclass(as.POSIXct(dates))
  47
  48     xx<-NULL;
  49     yy<-NULL;
  50
  51     for ( i in seq(1,length(hbreaks)-1) )
  52     {
  53         # get range from ns
  54         ns_sub <- ns[which(ns$date > hbreaks[i] & ns$date <= hbreaks[i+1] & ns$status == 'BOOT'),]
  55         nodes <- length(ns_sub$date)
  56
  57         xx<- c(xx, hbreaks[i])
  58         yy<- c(yy, nodes)
  59
  60     }
  61     m<- months[1:length(months)-1]
  62     return (rbind(xx,yy,m))
  63 }
  64
  65
  66 open_tickets <- function (t, tg)
  67 {
  68     xx<-NULL;
  69     yy<-NULL;
  70
  71     hbreaks<-tg$day_ts
  72
  73     for ( i in seq(1,length(hbreaks)-1) )
  74     {
  75         # identify any tickets with a start time in range, lastreply in range
  76         # or where both start is less and lastreply is greater than the range
  77         t_sub <- t[which( (t$start < hbreaks[i] & t$lastreply > hbreaks[i+1]) |
  78                           (t$start > hbreaks[i] & t$start <= hbreaks[i+1]) |
  79                           (t$lastreply > hbreaks[i] & t$lastreply <= hbreaks[i+1]) ),]
  80         tickets <- length(t_sub$start)
  81
  82         xx<- c(xx, hbreaks[i])
  83         yy<- c(yy, tickets)
  84     }
  85     return (rbind(xx,yy))
  86 }
  87
  88 online_nodes <- function (fb)
  89 {
  90     breaks <- unique(fb$timestamp)
  91     n<-NULL
  92     o<-NULL
  93     x<-NULL
  94     for (i in seq(1,length(breaks)) )
  95     {
  96         ts <- breaks[i]
  97         sub <- fb[which(fb$timestamp == ts),]
  98         node_count   <- length(unique(sub$hostname))
  99         online_count <- length(unique(sub$hostname[which(sub$state=='BOOT')]))
 100         x<-c(x,ts)
 101         n<-c(n,node_count)
 102         o<-c(o,online_count)
 103     }
 104     print(length(x))
 105     print(length(n))
 106     print(length(o))
 107     return (rbind(x,n,o))
 108 }
 109
 110 #####
 111
 112 # system("rt_s1_raw_dump.py --runsql");
 113 # system("rt_s2_parse_raw.py 3 > rt_data.csv");
 114 # t <- read.csv('rt_data_2004-2011.csv', sep=',', header=TRUE)
 115 #t <- read.csv(, sep=',', header=TRUE)
 116
 117 draw_rt_data <- function (input_filename, output_filename, start_date, end_date, draw=TRUE, one=FALSE)
 118 {
 119     t <- read.csv(input_filename, sep=',', header=TRUE)
 120     t2 <- t[which(t$complete == 1),]
 121
 122     tg <- time_graph_setup(start_date, end_date)
 123     ot <- open_tickets(t2, tg)
 124
 125     if ( draw == TRUE ) {
 126         start_image(output_filename, width=600, height=400)
 127     }
 128     if ( one == TRUE )
 129     {
 130         par(mfrow=c(1,1))
 131         par(mai=c(0.8,1,0.4,0.1))
 132     } else {
 133         par(mfrow=c(2,1))
 134         par(mai=c(0,1,0.3,0.1))
 135     }
 136
 137     x1<-as.numeric(ot[1,])
 138     y1<-as.numeric(ot[2,])
 139
 140     a_ot<-lowess_smooth(x1, y1)
 141
 142     plot(x1, y1, col='grey80', type='l', axes=F,
 143         ylab="a) Open Tickets (tickets/day)", xlab="Date",
 144         ylim=c(0,120)) # , ylim=c(0,260))
 145     lines(a_ot$x, round(a_ot$y), col='black')
 146
 147     axis(2, las=1)
 148     if ( one == TRUE ) {
 149         axis(1, labels=tg$month_str, at=tg$month_ts, cex.axis=0.7)
 150         axis(1, labels=tg$year_str, at=tg$year_ts, cex.axis=0.7, line=1, lwd=0)
 151     }
 152
 153
 154     abline(h=15, lty=3, col='grey80')
 155     abline(h=25, lty=3, col='grey80')
 156     abline(h=40, lty=3, col='grey80')
 157
 158     plc_releases(120)
 159     if ( one == FALSE )
 160     {
 161         par(mai=c(1,1,0.1,0.1))
 162         for ( s in c(5) )
 163         {
 164             d <- median_time_to_resolve_window(t2, tg, s) # "2004/1/1", "2011/1/28", s, "%b")
 165             plot(d[,1], exp(as.numeric(d[,5]))/24, type='l', lty=1, xlab="",
 166                     axes=F, ylim=c(0.01, 15), ylab="b) Resolution Time by", col='black',
 167                     xlim=c(min(x1), max(x1)))
 168             mtext("Quartile (days)", 2, 2)
 169             lines(d[,1], exp(as.numeric(d[,4]))/24, lty=1, col='grey50')
 170             lines(d[,1], exp(as.numeric(d[,3]))/24, lty=1, col='grey75')
 171             axis(1, labels=tg$month_str, at=tg$month_ts, cex.axis=0.7)
 172             axis(1, labels=tg$year_str, at=tg$year_ts, cex.axis=0.7, line=1, lwd=0)
 173             axis(2, labels=c(0,1,4,7,14), at=c(0,1,4,7,14), las=1)
 174             m<-round(max(exp(as.numeric(d[,4]))/24), 2)
 175         }
 176
 177         abline(h=1, lty=3, col='grey80')
 178         abline(h=4, lty=3, col='grey80')
 179         abline(h=7, lty=3, col='grey80')
 180
 181         planetlab_releases(15)
 182     }
 183
 184     if ( draw == TRUE ) {
 185         end_image()
 186     }
 187 }
 188
 189 #system("./rt_s2_parse_raw.py 3 > rt_data_2004-2011.csv");
 190 draw_rt_data('rt_data_2004-2011.csv', "rt_operator_support_2004-2011.png", "2004/1/1", "2011/6/1", TRUE, TRUE)
 191 #draw_rt_data('rt_data_monitor_2004-2011.csv',"rt_operator_monitor_2004-2011.png", "2004/1/1", "2011/4/1")
 192
 193 #draw_rt_data('short_support_20110101.csv',"rt_short_2011.png", "2010/11/1", "2011/4/1", FALSE)