-
Notifications
You must be signed in to change notification settings - Fork 2
/
results.R
140 lines (114 loc) · 4.54 KB
/
results.R
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
library(tidyverse)
source('analyse.R')
load('result_data.RData')
# -- Helper Functions
tidylatency <- function(test_log){
data.frame(
latency = test_log %>% pull(latency_sec),
users = test_log %>% pull(tar_sess) %>% unique(),
stringsAsFactors = FALSE
)
}
prepare <- function(test_logs, ratio = 1) {
all <- map_df(test_logs, tidylatency)
if (ratio > 1) {
all$r_procs <- all$users / ratio
} else {
all$r_procs <- 1
}
grouping <- as.symbol(ifelse(ratio > 1, 'r_procs', 'users'))
all %>%
group_by(!!grouping) %>%
summarise(avg = mean(latency),
med = median(latency),
upper = quantile(latency, 0.75),
lower = quantile(latency, 0.25)) %>%
right_join(all)
}
latency_plot <- function(prepped, title, grouping) {
#facet <- paste0(grouping, '~ .')
facet <- grouping
ggplot(prepped) +
geom_density(aes_string(group = grouping, x = "latency")) +
geom_point(aes_string(group = grouping, x = "avg"), y = 0.5) +
geom_point(aes_string(group = grouping, x = "med"), y = 0.5, color = 'red') +
facet_wrap(facet,
labeller = function(l){map_df(l, ~paste0(as.character(.x), ' ', grouping))},
ncol = 1
) +
scale_y_continuous(breaks = NULL) +
labs(
title = NULL,
caption = 'Red - Median; Black - Mean',
x = 'Latency (Secs)',
y = NULL
) +
theme(
text = element_text(size = 20)
)
}
# -- First, look at the app without Caching
# Number of Users for 1 R process
# o_40_1 <- analyse('old_tests/40_1', '40', 'old_tests/30min.txt', 39)
# o_30_1 <- analyse('old_tests/30_1', '30', 'old_tests/30min.txt', 28)
# o_20_1 <- analyse('old_tests/20_1', '20', 'old_tests/30min.txt', 20)
nocache_single <- prepare(list(o_20_1, o_30_1,o_40_1))
latency_plot(nocache_single, 'Latency vs # of Users for 1 R process', 'users')
# Linear Scaling
# o_60_2 <- analyse('old_tests/60_2', '60', 'old_tests/30min.txt', 60 )
# o_90_3 <- analyse('old_tests/90_3', '90', 'old_tests/30min.txt', 90 )
# o_120_4 <- analyse('old_tests/120_4', '120', 'old_tests/30min.txt', 120 )
# o_150_5 <- analyse('old_tests/150_5', '150', 'old_tests/30min.txt', 150)
# o_180_6 <- analyse('old_tests/180_6', '180', 'old_tests/30min.txt', 180)
# o_210_7 <- analyse('old_tests/210_7', '210', 'old_tests/30min.txt', 210)
# o_240_8 <- analyse('old_tests/240_8', '240', 'old_tests/30min.txt', 240)
nocache_many <- prepare(list(o_30_1, o_60_2,o_90_3,o_120_4,o_150_5, o_180_6, o_210_7, o_240_8), ratio = 30)
latency_plot(nocache_many, 'Nearly Linear Scaling: Latency vs # of R Procs (30 Users / Proc)', 'users')
# Some linear scaling plots
nocache_many %>%
select(users, Median = med, Average = avg, `75th Percentile` = upper, `25th Percentile` = lower) %>%
unique() %>%
ggplot() +
geom_line(aes(users, Median), color = 'red') +
geom_line(aes(users, Average), color = 'black') +
geom_errorbar(aes(users, ymax = `75th Percentile`, ymin = `25th Percentile`), alpha = 0.35) +
labs(
title = NULL,
x = 'Users',
y = 'Latency (Secs)'
) +
theme_minimal() +
theme(text = element_text(size = 20))
data.frame(
r_processes = 1:8,
users = 540*1:8
) %>%
ggplot() +
geom_line(aes(r_processes, users)) +
labs(
title = NULL,
x = 'R Processes',
y = 'Users'
) +
theme_minimal() +
theme(text = element_text(size = 20))
# -- Next, look at the same metrics, but for the cache optimized app
# single process
# c_90_1 <- analyse('90_1', '90', '15min.log', 90)
# c_60_1 <- analyse('60_1', '60', '15min.log', 60)
# c_120_1 <- analyse('120_1', '120', '15min.log', 120 )
cache_single <- prepare(list(c_60_1,c_90_1,c_120_1))
latency_plot(cache_single, 'Latency vs # of Users for 1 R process, 3x on Cache Optimized App', 'users')
# across processess
# c_180_2 <- analyse('180_2', '180', '15min.log',170)
# c_360_4 <- analyse('360_4', '360', '15min.log',360)
# c_540_6 <- analyse('540_6-2', '540', '15min.log',540)
# c_630_7 <- analyse('630_7', '630', '15min.log',600)
# c_720_8 <- analyse('720_8', '720', '15min.log',720)
cache_many <- prepare(list(c_180_2, c_360_4, c_540_6, c_630_7, c_720_8), 90)
latency_plot(cache_many, 'Not-so Linear Scaling: Latency vs # of R Procs (90 Users / Proc)', 'r_procs')
# across nodes
# c_2700_30 <- analyse('2700_30', '2700', '15min.log', 2700)
# c_1080_12 <- analyse('1080_12', '1080', '15min.log', 1080)
cache_nodes <- prepare(list(c_540_6, c_1080_12, c_2700_30), 90)
latency_plot(cache_nodes, 'Linear Scaling Across Nodes (540 Users / Node)', 'users')