viz script: only keep relevant parts /artifacts

scripts-r/viz-campaign.R

 #!/usr/bin/env Rscript
 library(tidyverse)
-library(GGally)
 library(viridis)
-agg_file = '/tmp/agg-result.csv'
+agg_file = './expe1/agg-result.csv'
 output_dir = '/tmp'
 epoch_m100 = ymd_hms('2022-01-01 00:00:00')
 `%notin%` = Negate(`%in%`)
@@ -91,7 +90,7 @@ ggsave(sprintf("%s/utilization-against-powercap-predictor.pdf", output_dir), wid
 
 # mean turnaround time metrics
 data_nz %>% ggplot(aes(x=powercap_dynamic_value_ratio, y=mean_turnaround_time_minus_zero, color=predictor_name)) +
-  geom_jitter(width=1/100, height=0) +
+  geom_point() +
   geom_smooth(method = "lm", se = FALSE) +
   geom_hline(yintercept=0) +
   theme_bw() +
@@ -104,58 +103,17 @@ data_nz %>% ggplot(aes(x=powercap_dynamic_value_ratio, y=mean_turnaround_time_mi
   expand_limits(x=0) +
   labs(
     y="Mean turnaround time difference against EASY without any powercap for each simulation",
-    x="Powercap value (proportion of the maximum dynamic power range). Shown with horizontal jitter."
+    x="Powercap value (proportion of the maximum dynamic power range)"
   )
 ggsave(sprintf("%s/mean-turnaround-time-against-powercap-predictor.pdf", output_dir), width=16, height=9)
 
-# mean waiting time
-data_nz %>% ggplot(aes(x=powercap_dynamic_value_ratio, y=mean_waiting_time_minus_zero, color=predictor_name)) +
-  geom_jitter(width=1/100, height=0) +
-  geom_smooth(method = "lm", se = FALSE) +
-  geom_hline(yintercept=0) +
-  theme_bw() +
-  theme(legend.position='top', legend.title=element_blank()) +
-  guides(color = guide_legend(nrow = 1)) +
-  scale_x_continuous(breaks=seq(0,0.7,0.1), labels = scales::percent) +
-  scale_y_continuous() +
-  scale_color_viridis(discrete=TRUE) +
-  facet_wrap(vars(start_dt_s), scales='free_y') +
-  expand_limits(x=0) +
-  labs(
-    y="Mean waiting time difference against EASY without any powercap for each simulation",
-    x="Powercap value (proportion of the maximum dynamic power range). Shown with horizontal jitter."
-  )
-ggsave(sprintf("%s/mean-waiting-time-against-powercap-predictor.pdf", output_dir), width=16, height=9)
-
-
-# power (mean+p1+p99) for all instances, facetted by workload, colored by predictor
 max_observed_total_power = 955080
 max_power_per_node = 2100.0
 min_power_per_node = 240.0
 nb_nodes = 980
 max_dynamic_power = max_observed_total_power - min_power_per_node * nb_nodes
-data_nz %>% ggplot(aes(x=powercap_dynamic_value_ratio)) +
-  #geom_jitter(width=1/100, height=0) +
-  #geom_smooth(method = "lm", se = FALSE) +
-  geom_ribbon(aes(ymin=power_p1/max_dynamic_power, ymax=power_p99/max_dynamic_power, fill=predictor_name), alpha=0.1) +
-  geom_line(aes(y=mean_power/max_dynamic_power, color=predictor_name)) +
-  geom_line(aes(y=power_p1/max_dynamic_power, color=predictor_name), linetype='dotted', show.legend = FALSE) +
-  geom_line(aes(y=power_p99/max_dynamic_power, color=predictor_name), linetype='dotted', show.legend = FALSE) +
-  geom_abline(slope=1) +
-  theme_bw() +
-  theme(legend.position='top', legend.title=upper_bound) +
-  scale_color_viridis(discrete=TRUE) +
-  scale_fill_viridis(discrete=TRUE) +
-  facet_wrap(vars(start_dt_s)) +
-  expand_limits(x=0) +
-  labs(
-    y="Power consumption during the powercap-constrained 3-hour time window (proportion of the maximum dynamic power range).",
-    x="Powercap value (proportion of the maximum dynamic power range)."
-  )
-ggsave(sprintf("%s/power-consumption-mean-p1-p99-against-powercap-predictor.pdf", output_dir), width=16, height=9)
 
 # distribution of the mean power used for each predictor
-#powercap_ratios_values_to_show = c(0.1, 0.25, 0.4, 0.55, 0.7)
 powercap_ratios_values_to_show = c(0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7)
 width_scale=0.3
 data_nz %>%
@@ -211,7 +169,6 @@ data_nz %>%
 scale=0.9
 ggsave(sprintf("%s/sched-mtt-distribution.pdf", output_dir), width=8*scale, height=4*scale)
 
-
 # show all columns...
 options(dplyr.width = Inf)
 
@@ -319,30 +276,14 @@ against_mean = data_p_mtt %>%
     all_above_mean = real_mean_above_mean & upper_bound_above_mean & real_max_above_mean & max_above_mean
   )
 
-
-# overview of whether each predictor breaks the powercap
-data %>%
-  mutate(breaks_powercap = surplus_energy > 0) %>%
-  group_by(predictor_name, breaks_powercap) %>%
-  summarize(occ=n())
-
-
-
-data_nz %>% filter(start_dt_s != 18474670) %>% ggplot(aes(x=predictor_name, y=mean_waiting_time_minus_zero)) +
-  geom_violin() +
-  geom_jitter(alpha=0.1) +
-  geom_boxplot(width=1/8, outlier.shape=NA) +
-  facet_wrap(vars(start_dt_s)) +
-  theme_bw()
-
 data_nz %>% filter(start_dt_s != 18474670) %>% ggplot(aes(y=energy_from_powercap / 1e9, x=mean_utilization / 980, color=predictor_name)) +
   geom_point() +
   stat_ellipse() +
   theme_bw() +
-  #facet_wrap(vars(powercap_dynamic_value_ratio)) +
   scale_color_viridis(discrete=TRUE)
+ggsave(sprintf("%s/meh1.pdf", output_dir), width=16, height=9)
 
-data %>% ggplot() +
+data_nz %>% ggplot() +
   geom_violin(aes(x=predictor_name, y=energy_from_powercap / 1e9)) +
   geom_jitter(aes(x=predictor_name, y=energy_from_powercap / 1e9), alpha=0.1) +
   geom_boxplot(aes(x=predictor_name, y=energy_from_powercap / 1e9), width=0.025, outlier.shape=NA) +
@@ -351,8 +292,9 @@ data %>% ggplot() +
     x="Power predictor",
     y="Distribution of the energy consumed during the constrained period for each simulation (GJ).\nComputed as the integral of the dynamic power minus the dynamic powercap value."
   )
+ggsave(sprintf("%s/meh2.pdf", output_dir), width=16, height=9)
 
-data %>% ggplot() +
+data_nz %>% ggplot() +
   geom_violin(aes(x=predictor_name, y=surplus_energy / 1e9)) +
   geom_jitter(aes(x=predictor_name, y=surplus_energy / 1e9), alpha=0.1) +
   geom_boxplot(aes(x=predictor_name, y=surplus_energy / 1e9), width=0.025, outlier.shape=NA) +
@@ -361,8 +303,9 @@ data %>% ggplot() +
     x="Power predictor",
     y="Distribution of the surplus energy consumed during the constrained period for each simulation (GJ).\nComputed as the integral of the dynamic power minus the dynamic powercap value, only keeping positive values."
   )
+ggsave(sprintf("%s/meh3.pdf", output_dir), width=16, height=9)
 
-data %>% ggplot() +
+data_nz %>% ggplot() +
   geom_violin(aes(x=predictor_name, y=unused_energy / 1e9)) +
   geom_jitter(aes(x=predictor_name, y=unused_energy / 1e9), alpha=0.1) +
   geom_boxplot(aes(x=predictor_name, y=unused_energy / 1e9), width=0.025, outlier.shape=NA) +
@@ -371,3 +314,4 @@ data %>% ggplot() +
     x="Power predictor",
     y="Distribution of the unused energy consumed during the constrained period for each simulation (GJ).\nComputed as the integral of the dynamic power minus the dynamic powercap value, only keeping negative values."
   )
+ggsave(sprintf("%s/meh4.pdf", output_dir), width=16, height=9)