Shinyappsio Usage and projected needs

Looking at pas 90 days of Shinyappsio consumption to inform 2023 renewal.

Executive take-aways

• High consumption for bchc_covid19 app
  • We are still seeing quite a bit of traffic here (see supplemntal figure below)
  • But consumption is driven primarily by some residual settings that were not turned off.
  • We removed the 24/7 on-time but…
    
  • I forgot there were some app settings tweaked for persistence that was causing high consumption. e.g. Instance Idle Timeout Time after which your idle application will be stopped. The min is 15 minutes but it was set to 60 minutes.
  • We set all settings to default… so we should expect much lower consumption from this app
• Overall consumption is ~500 hours per month
• Taking into account the projected decrease in bchc_covid19, we expect much less than 500 hours per month of usage with our current apps
• A basic plan ($440/year) with 500 hours per month would cover our needs.

See below for supporting materials.

Plots

Quick EDA on ShinyAppsio usage to inform plan renewal for 2023. We’ll look at by app daily vs monthly consumption as well as account level monthly consumption.

Lets first pull in some usage data!

Code

## Dependencies
library(rsconnect)
library(tidyverse)
library(purrr)
library(lubridate)
library(highcharter)

## Get App names
apps = applications(account = "drexel-uhc",
                    server="shinyapps.io") %>% 
  as_tibble() %>% 
  select(id, name) %>% 
  mutate(id = as.character(id))

### Pull Data
start_date = "2022-10-03"
end_date = Sys.Date()

api_result = rsconnect::accountUsage(
  account = "drexel-uhc",
  server="shinyapps.io",
  usageType = "hours",
  from = start_date,
  until  = end_date,
  interval = "1d"
)
.x = api_result$points[[1]]
list_result = api_result$points %>% 
  map(  ~{
    hours_tmp = map(.x,2)%>% 
      unlist() 
    tibble(date = seq(ymd(start_date),
                      ymd(end_date)-1,
                      by = "day") )  %>% 
      mutate(day = row_number()) %>% 
      mutate(hours = hours_tmp)
  }) 
df_result = map2_df(list_result,names(list_result),
                    ~{.x %>% mutate(id_chr = .y)}) %>% 
  mutate(id  = str_remove(id_chr,"application-")) %>% 
  left_join(apps) %>% 
  select(name, date, day, hours) %>% 
  arrange(name,date) #%>% 
  # group_by(name) %>% 
  # group_modify(~.x %>% mutate(hours_cum = cumsum(hours))) %>% 
  # ungroup()  %>%
# mutate(month = month) %>% 
# select(month, everything())

df_usage_daily = df_result %>% 
  filter(!is.na(name)) %>% 
  rename(app = name) %>% 
  mutate(month = month(date),
         year = year(date)) %>% 

  arrange(app, year, month)


df_useage_monthly = df_usage_daily %>% 
  filter(month != "2") %>% 
  group_by(app, month, year) %>% 
  summarise(hours = sum(hours)) %>% 
  ungroup()%>% 
  arrange(app, year, month) %>% 
  mutate(date = ymd(paste0(year,"/",month,'/15')))

df_usage_monthly_overall = df_usage_daily %>% 
  filter(month != "2") %>% 
  group_by(month, year) %>% 
  summarize(hours = sum(hours)) %>% 
  mutate(date = ymd(paste0(year,"/",month,'/15')))

By app

Note that in highcharted you can click the legend to remove or add back specific lines. or exmaple you can click bchc_covid19 to remove it and rescale the plot.

Code

df_usage_daily %>%
  hchart("line",
         hcaes(x = date, y = hours, group = app)) %>% 
  hc_title(text = 'Daily trends over last 90 days by app')

Code

df_useage_monthly %>%
  filter(month != "2") %>% 
  hchart("line",
         hcaes(x = date, y = hours, group = app)) %>% 
  hc_title(text = 'Monthly trends over last 90 days by app') %>% 
  hc_subtitle(text = 'Monthly cumulative hours used is displayed at the midpoint of each month. ')

Overall

Code

df_usage_monthly_overall %>%
  arrange(date) %>% 
  hchart("line",
         hcaes(x = date, y = hours)) %>% 
  hc_title(text = 'Overall usage over 3 months') %>% 
  hc_subtitle(text = 'Monthly cumulative hours used is displayed at the midpoint of each month.') %>% 
  hc_yAxis(min = 0)

Supplemental figures