In order to test, measure, and evaluate using client-side sampling to reduce the costs of GLAM, the following metrics have been chosen to instrument in duplicate with a side-by-side metric that is disabled by default. We will then use an experiment to enable these metrics for the same percentage of the population that use for server-side sampling of GLAM data. The metrics proposed for side-by-side instrumentation are from different segments of the population to ensure low reporting metrics are as reliable as high reporting metrics, and include 2 timing distributions and 2 counters from each segment. The initial metrics that have been selected include: Low population: - [power.cpu_time_bogus_values](https://dictionary.telemetry.mozilla.org/apps/firefox_desktop/metrics/power_cpu_time_bogus_values) - [javascript.pageload.protect_time](https://dictionary.telemetry.mozilla.org/apps/firefox_desktop/metrics/javascript_pageload_protect_time) - [networking.http_content_html5parser_ondatafinished_to_onstop_delay](https://dictionary.telemetry.mozilla.org/apps/firefox_desktop/metrics/networking_http_content_html5parser_ondatafinished_to_onstop_delay) - [networking.os_socket_limit_reached](https://dictionary.telemetry.mozilla.org/apps/firefox_desktop/metrics/networking_os_socket_limit_reached) Med population: - [network.sub_complete_load_net](https://dictionary.telemetry.mozilla.org/apps/firefox_desktop/metrics/network_sub_complete_load_net) - [downloads.panel_shown](https://dictionary.telemetry.mozilla.org/apps/firefox_desktop/metrics/downloads_panel_shown) - [pdfjs.used](https://dictionary.telemetry.mozilla.org/apps/firefox_desktop/metrics/pdfjs_used) - [performance.pageload.async_sheet_load](https://dictionary.telemetry.mozilla.org/apps/firefox_desktop/metrics/performance_pageload_async_sheet_load) High population: - [perf.largest_contentful_paint](https://dictionary.telemetry.mozilla.org/apps/firefox_desktop/metrics/perf_largest_contentful_paint) - [browser.engagement.active_ticks](https://dictionary.telemetry.mozilla.org/apps/firefox_desktop/metrics/browser_engagement_active_ticks) - [performance.clone.deserialize.time](https://dictionary.telemetry.mozilla.org/apps/firefox_desktop/metrics/performance_clone_deserialize_time) - [power.total_cpu_time_ms](https://dictionary.telemetry.mozilla.org/apps/firefox_desktop/metrics/power_total_cpu_time_ms) I will begin reaching out to owners of these metrics to ensure that there will be no problems with our adding parallel instrumentation before proceeding, and will use this bug to track progress and patches related to implementation.
Bug 1947604 Comment 0 Edit History
Note: The actual edited comment in the bug view page will always show the original commenter’s name and original timestamp.
In order to test, measure, and evaluate using client-side sampling to reduce the costs of GLAM, the following metrics have been chosen to instrument in duplicate with a side-by-side metric that is disabled by default. We will then use a Nimbus rollout to enable these metrics for a similar percentage of the population that use for server-side sampling of GLAM data. The metrics proposed for side-by-side instrumentation are from different segments of the population to ensure low reporting metrics are as reliable as high reporting metrics, and include 2 timing distributions and 2 counters from each segment. The initial metrics that have been selected include: Low population: - [power.cpu_time_bogus_values](https://dictionary.telemetry.mozilla.org/apps/firefox_desktop/metrics/power_cpu_time_bogus_values) - [javascript.pageload.protect_time](https://dictionary.telemetry.mozilla.org/apps/firefox_desktop/metrics/javascript_pageload_protect_time) - [networking.http_content_html5parser_ondatafinished_to_onstop_delay](https://dictionary.telemetry.mozilla.org/apps/firefox_desktop/metrics/networking_http_content_html5parser_ondatafinished_to_onstop_delay) - [networking.os_socket_limit_reached](https://dictionary.telemetry.mozilla.org/apps/firefox_desktop/metrics/networking_os_socket_limit_reached) Med population: - [network.sub_complete_load_net](https://dictionary.telemetry.mozilla.org/apps/firefox_desktop/metrics/network_sub_complete_load_net) - [downloads.panel_shown](https://dictionary.telemetry.mozilla.org/apps/firefox_desktop/metrics/downloads_panel_shown) - [pdfjs.used](https://dictionary.telemetry.mozilla.org/apps/firefox_desktop/metrics/pdfjs_used) - [performance.pageload.async_sheet_load](https://dictionary.telemetry.mozilla.org/apps/firefox_desktop/metrics/performance_pageload_async_sheet_load) High population: - [perf.largest_contentful_paint](https://dictionary.telemetry.mozilla.org/apps/firefox_desktop/metrics/perf_largest_contentful_paint) - [browser.engagement.active_ticks](https://dictionary.telemetry.mozilla.org/apps/firefox_desktop/metrics/browser_engagement_active_ticks) - [performance.clone.deserialize.time](https://dictionary.telemetry.mozilla.org/apps/firefox_desktop/metrics/performance_clone_deserialize_time) - [power.total_cpu_time_ms](https://dictionary.telemetry.mozilla.org/apps/firefox_desktop/metrics/power_total_cpu_time_ms) I will begin reaching out to owners of these metrics to ensure that there will be no problems with our adding parallel instrumentation before proceeding, and will use this bug to track progress and patches related to implementation.