deletions | additions
diff --git a/data/plot.R b/data/plot.R
index 1de797f..c862fbf 100755
--- a/data/plot.R
+++ b/data/plot.R
...
#!/usr/bin/env Rscript
source('common.R')
capply <- function(col, func) unlist(lapply(col, func))
data <- function(d) {
d$abort_rate <- d$txn_failed / (d$txn_count + d$txn_failed)
d$throughput <- d$txn_count * num(d$nclients) / d$total_time
# d$throughput <- d$ntxns * num(d$nclients) / d$total_time
d$avg_latency_ms <- d$txn_time / d$txn_count * 1000
d$Commutativity d$`Concurrency Control` <- revalue(d$ccmode, c(
'bottom'='base (none)',
'simple'='simple' 'rw'='reader/writer',
'simple'='commutative'
))
d$Graph <- capply(d$gen, function(s) gsub('kronecker:.+','kronecker',s))
d$facet <- sprintf('sh: %d, u: %d, g: %s', num(d$nshards), d$initusers, d$Graph)
d$gen_label <- sprintf('%d users\n%s', d$initusers, d$Graph)
return(d)
}
d.all <- data(db("
select * from tapir where
generator_time is not null and total_time is not null
and (initusers = 50 or initusers = 500)
and name like 'claret-v%'
",
factors=c('nshards', 'nclients'),
...
d <- data(db("
select * from tapir where
generator_time is not null and total_time is not null
and
(initusers = 50 or initusers = 500)
and name like
'claret-v0.3%' 'claret-v0.5.1%'
and ccmode != 'bottom'
",
factors=c('nshards', 'nclients'),
numeric=c('total_time', 'txn_count')
...
theme_mine
, name='throughput_compare_versions', w=6, h=7)
d.u <- subset(d,
(initusers == 4096 | initusers ==
500 512) &
gen nshards ==
'uniform') 4)
save(
ggplot(d.u, aes(
x = nclients,
y = throughput,
group =
Commutativity, `Concurrency Control`,
fill =
Commutativity, `Concurrency Control`,
color =
Commutativity `Concurrency Control`
))+
# geom_meanbar()+
stat_smooth()+
common_layers facet_wrap(~gen_label)+
theme_mine
, name='throughput', w=4,
h=3) h=5)
save(
ggplot(d, aes(
x = nclients,
y = throughput,
group = `Concurrency Control`,
fill = `Concurrency Control`,
color = `Concurrency Control`
))+
# geom_meanbar()+
stat_smooth()+
# facet_grid(nshards~initusers, labeller=label_pretty)+
facet_wrap(~facet)+
theme_mine
, name='throughput_explore', w=8, h=7)
save(
ggplot(d.u, aes(
x = nclients,
y = avg_latency_ms,
group =
Commutativity, `Concurrency Control`,
fill =
Commutativity, `Concurrency Control`,
color =
Commutativity `Concurrency Control`
))+
stat_smooth()+
geom_hline(y=0)+
common_layers facet_wrap(~gen_label)+
theme_mine
, name='avg_latency', w=4,
h=3) h=5)
save(
ggplot(d, aes(
x = nclients,
y = avg_latency_ms,
group =
Commutativity, `Concurrency Control`,
fill =
Commutativity, `Concurrency Control`,
color =
Commutativity `Concurrency Control`
))+
# geom_meanbar()+
# stat_summary(fun.y='mean', geom='bar', position='dodge')+
...
facet_grid(nshards~initusers, labeller=label_pretty)
, name='avg_latency_explore', w=8, h=6)
subset(d.u, select=c('nshards','nclients','Graph','Concurrency Control','abort_rate','throughput'))
save(
ggplot(d.u, aes(
x = nclients,
y = abort_rate,
group =
Commutativity, `Concurrency Control`,
fill =
Commutativity, `Concurrency Control`,
color =
Commutativity `Concurrency Control`
))+
stat_smooth()+
common_layers+
geom_hline(y=0)+
facet_grid(~nshards, labeller=label_pretty) facet_wrap(~gen_label)+
theme_mine
, name='abort_rates', w=4,
h=3) h=5)
save(
ggplot(d, aes(
x = nclients,
y = abort_rate,
group =
ccmode, `Concurrency Control`,
fill =
ccmode, `Concurrency Control`,
color =
ccmode `Concurrency Control`
))+
# geom_meanbar()+
# stat_summary(fun.y='mean', geom='bar', position='dodge')+
stat_smooth()+
common_layers+
# geom_hline(y=0)+
facet_grid(nshards~initusers, labeller=label_pretty) facet_wrap(~facet)
, name='abort_rates_exploration',
w=4, h=3) w=7, h=7)
d$op_retries_total <- d$op_retries * num(d$nclients)
d$op_retry_ratio <- d$op_retries / d$op_count
save(
ggplot(d, ggplot(subset(d), aes(
x = nclients,
y = op_retry_ratio,
group =
ccmode, `Concurrency Control`,
fill =
ccmode, `Concurrency Control`,
color =
ccmode `Concurrency Control`
))+
stat_smooth()+
common_layers+
geom_hline(y=0)+
# facet_grid(nshards~initusers, labeller=label_pretty)
facet_wrap(~facet)
, name='op_retries', w=8, h=6)
d.u$retwis_newuser_latency <- d.u$retwis_newuser_time / d.u$retwis_newuser_count
d.u$retwis_post_latency <- d.u$retwis_post_time / d.u$retwis_post_count
d.u$retwis_repost_latency <- d.u$retwis_repost_time / d.u$retwis_repost_count
d.u$retwis_timeline_latency <- d.u$retwis_timeline_time / d.u$retwis_timeline_count
d.u$retwis_follow_latency <- d.u$retwis_follow_time / d.u$retwis_follow_count
save(
ggplot(d.u, aes(
x = nclients,
y = retwis_repost_latency * 1000,
group = `Concurrency Control`,
color = `Concurrency Control`,
fill = `Concurrency Control`
))+
stat_smooth()+
facet_wrap(~Graph)+
theme_mine
, name='repost_txn_latency', w=4, h=3)
d.lat <- melt(d.u,
measure=c(
'retwis_newuser_latency',
'retwis_post_latency',
'retwis_repost_latency',
'retwis_timeline_latency',
'retwis_follow_latency'
)
)
d.lat$txn_type <- capply(d.lat$variable, function(s) gsub('retwis_(\\w+)_latency','\\1', s))
d.lat$latency_ms <- d.lat$value * 1000
save(
ggplot(d.lat, aes(
x = nclients,
y = latency_ms,
group = txn_type,
color = txn_type,
fill = txn_type
))+
stat_smooth()+
facet_wrap(~Graph)+
theme_mine
, name='txn_breakdown_latency', w=4, h=3)
d.ct <- melt(d.u,
measure=c(
'retwis_newuser_count',
'retwis_post_count',
'retwis_repost_count',
'retwis_timeline_count',
'retwis_follow_count'
)
)
d.ct$txn_type <- capply(d.ct$variable, function(s) gsub('retwis_(\\w+)_count','\\1', s))
d.ct$total_count <- d.ct$value * num(d.ct$nclients)
save(
ggplot(d.ct, aes(
x = txn_type,
y = total_count,
group = txn_type,
fill = txn_type
))+
geom_meanbar()+
facet_wrap(~Graph)+
theme_mine
, name='txn_counts', w=4, h=3)
d.m <- melt(subset(d, ccmode == 'simple'),
measure=c(
'retwis_newuser_success',
...
# geom_meanbar()+
stat_summary(fun.y='mean', geom='smooth')+
common_layers+
facet_grid(nshards~initusers, labeller=label_pretty) facet_wrap(~facet)
, name='txn_breakdown', w=8, h=6)
