meetings.Rmd

---
title: "5 min. Zoom Test Trace: Meetings"
date: "`r Sys.Date()`"
output:
  html_document:
    df_print: paged
params:
  input_file: "meetings.csv"
  fig_path: "fig/"
---

```{r setup, include=FALSE}
source("setup.R")

knitr::opts_chunk$set(dev = c("png", "pdf"), fig.width = 7, fig.height = 4,
  fig.align = "center", fig.keep = "high", fig.path = params$fig_path)

colors <- c("#22528E", "#882111", "#3D8926")
```

```{r import}
meetings <- read_csv(params$input_file, show_col_types = FALSE) %>%
  mutate(conn_type = as.factor(conn_type), zoom_type = as.factor(zoom_type))
```

### Summary

```{r summary}
list(
  streams        = meetings %>% nrow(),
  unique_streams = meetings$stream_id %>% unique() %>% length(),
  meetings       = meetings$meeting_id %>% unique() %>% length()
) %>% as_tibble_row()
```

```{r}
meetings %>%
  group_by(stream_id, zoom_type) %>%
  summarize(.groups="keep") %>%
  group_by(zoom_type) %>%
  summarize(count = n(), .groups="keep") %>%
  arrange(-count)
```

### Stream Duplicates

* limited to <= 20 duplicates

```{r stream-duplicates}
meetings %>%
  group_by(stream_id, zoom_type) %>%
  summarize(n = n(), .groups="keep") %>%
  filter(n > 1 & n <= 20) %>%
  ggplot(aes(x=factor(n))) +
    geom_bar(stat="count", aes(fill=zoom_type)) +
    scale_fill_manual(name="Type",values=colors, labels=c("Screen S.", "Audio", "Video")) +
    labs(x="# Stream Duplicates", y="Frequency") +
    theme_om()
```

### Meeting duration

```{r meeting-duration-cdf}
meetings %>%
  group_by(meeting_id) %>%
  summarize(start = min(start_ts_s), end = max(end_ts_s)) %>%
  mutate(duration_s = (end - start) / 60) %>%
  ggplot(aes(x=duration_s)) +
    labs(x="Meeting duration [min.]", y="CDF") +
    scale_x_continuous(breaks=seq(0,720,60)) +
    stat_ecdf(color=colors[1], size=0.75) +
    theme_om()
```

### Number of streams per meeting

```{r}
meetings %>%
  group_by(meeting_id) %>%
  summarize(streams = n()) %>%
  ggplot(aes(x=streams)) +
    labs(x="# Streams per meeting (log.)", y="CDF") +
    scale_x_log10() +
    annotation_logticks(sides="b", colour = "gray50") +
    stat_ecdf(color=colors[1], size=0.75) +
    theme_om()
```

### Number of unique streams per meeting

```{r}
meetings %>%
  group_by(meeting_id) %>%
  summarize(unique_streams = stream_id %>% unique() %>% length()) %>%
  ggplot(aes(x=unique_streams)) +
    labs(x="# Unique streams per meeting (log.)", y="CDF") +
    scale_x_log10() +
    annotation_logticks(sides="b", colour = "gray50") +
    stat_ecdf(color=colors[1], size=0.75) +
    theme_om()
```

### Number of participants per meeting

```{r}
participants_per_meeting <- meetings %>% 
  group_by(meeting_id) %>%
  arrange(ssrc, .by_group = TRUE) %>%
  mutate(participant = cumsum((ssrc - lag(ssrc)) > 1 & !is.na(lag(ssrc)))) %>%
  summarize(participants = participant %>% unique() %>% length())

participants_per_meeting %>%
  ggplot(aes(x=participants)) +
    labs(x="# Meeting Participants (log.)", y="CDF") +
    scale_x_log10() +
    annotation_logticks(sides="b", colour = "gray50") +
    stat_ecdf(color=colors[1], size=0.75) +
    theme_om()
```

```{r}
# group by meeting participant:
participants <- meetings %>%
  group_by(meeting_id) %>%
  arrange(ssrc, .by_group = TRUE) %>%
  mutate(participant_id = cumsum((ssrc - lag(ssrc)) > 1 & !is.na(lag(ssrc)))) %>%
  group_by(meeting_id, participant_id) %>%
  summarize(
    start_ts_s = min(start_ts_s),
    end_ts_s = max(end_ts_s),
    audio = 15 %in% unique(zoom_type),
    video = 16 %in% unique(zoom_type),
    screen = 13 %in% unique(zoom_type),
    start_ts_min = as.integer(min(start_ts_s) / 60), # 60s buckets
    end_ts_min = as.integer(max(end_ts_s) / 60),
    .groups="keep"
  )
```

```{r}
list(participants = participants %>% nrow()) %>% as_tibble_row()
```



```{r}
# generate time series from meeting/participant list:
participants_over_time <- participants %>%
  group_by(meeting_id, participant_id) %>%
  summarize(min_activity = seq(start_ts_min, end_ts_min, 1), .groups="keep") %>%
  group_by(min_activity) %>%
  summarize(
    meeting_count = meeting_id %>% unique() %>% length(),
    participant_count = unique(meeting_id, participant_id) %>% length(),
    .groups = "keep"
  ) %>%
  mutate(ts = as_datetime((min_activity - 4*60) * 60)) # LT = UTC-4
```


### Number of concurrent meetings on campus
  
```{r meeting-count-ts}
participants_over_time %>%
  ggplot(aes(x=ts, y=meeting_count)) +
    labs(x = "Local Time", y = "Concurrent Meeting Count") +
    scale_y_continuous(limits = c(0, 150)) +
    geom_line(color=colors[1], size=0.75) +
    theme_om()
```

### Number of concurrent participants on campus

```{r participant-count-ts}
participants_over_time %>%
  ggplot(aes(x=ts, y=participant_count)) +
    labs(x = "Local Time", y = "Concurrent Participant Count") +
    scale_y_continuous(limits = c(0, 250)) +
    geom_line(color=colors[1], size=0.75) +
    theme_om()
```

### Media usage per participant

```{r}
participants %>%
  group_by(audio, video, screen) %>%
  summarize(count = n(), frac = n() / participants %>% nrow(), .groups="keep") %>%
  arrange(-frac)
```

### Fraction of speaking time when not muted
* by number of packets

```{r frac-speaking, fig.width=4, fig.height=3.8}
meetings %>%
  filter(zoom_type == 15 & audio_112_pkts > 0) %>%
  mutate(
    audio_112_frac = audio_112_pkts / (audio_99_pkts + audio_112_pkts),
  ) %>%
  select(speaking = audio_112_frac) %>%
  ggplot(aes(x=speaking)) +
    scale_x_continuous(expand = c(0.035, 0)) +
    scale_y_continuous(expand = c(0.035, 0)) +
    stat_ecdf(color=colors[1], size=0.75) +
    labs(x="Fraction Speaking (by packet count)", y="CDF") +
    theme_om()
```