OLD | NEW |
(Empty) | |
| 1 # Speed Operations: Services |
| 2 |
| 3 TL: simonhatch@<br> |
| 4 Team: simonhatch@ benjhayden@, dtu@, eakuefner@<br> |
| 5 TPM: benhenry@<br> |
| 6 Mailing List: speed-services-dev@chromium.org<br> |
| 7 **[go/chrome-speed-services](https://goto.google.com/chrome-speed-services)** |
| 8 |
| 9 ## Mission |
| 10 The mission of the Chrome Speed Services team is to develop the tools and |
| 11 infrastructure necessary for teams to detect, help diagnose, and follow through |
| 12 on performance regressions in Chrome. We accomplish this through a set of |
| 13 interconnected tools, which process data from the Chrome Benchmarking Team and |
| 14 automate the Chrome Speed Releasing workflow: |
| 15 |
| 16 * The [Chrome Performance Dashboard](https://chromeperf.appspot.com) stores |
| 17 over 4 million performance timeseries, and monitors hundreds of thousands |
| 18 of them for performance regressions. It integrates with Chrome's bug |
| 19 tracker for follow-through. |
| 20 * Most of our automated builds take too long to get performance data on every |
| 21 revision, so we have [bisect bots](bisects.md) to narrow down regressions. |
| 22 These bots also support running performance try jobs on unsubmitted CLs. |
| 23 * Whether a performance test is run locally or in the lab, right now or last |
| 24 year, we want to make it easy to understand and share the results. We |
| 25 develop [data formats](https://github.com/catapult-project/catapult/blob/mas
ter/docs/histogram-set-json-format.md) |
| 26 and user interfaces that make it easy to dig deeply into performance |
| 27 regressions. |
| 28 |
| 29 Here is a diagram of how our tooling fits into the chromium performance |
| 30 continuous build: |
| 31  |
| 32 |
| 33 |
| 34 ## 2017 Roadmap |
| 35 * **Improve our ability to bisect to root causes**. |
| 36 [Pinpoint](https://docs.google.com/document/d/1FKPRNU2kbPJ15p6XHO0itCjYtfvCp
Gt2IHblriTX1tg/edit) |
| 37 is a rewrite of the existing bisect system, design to address the |
| 38 shortcomings of a fully automated system running on buildbot. Tighter |
| 39 integration with the dashboard will allow better code sharing, more |
| 40 thorough test coverage, and a greatly improved and interactive user |
| 41 interface. |
| 42 * **Enable deep debugging and better numerics across our products.** |
| 43 We are working to make the dashboard, pinpoint, and telemetry all use our ne
w |
| 44 [HistogramSet](https://github.com/catapult-project/catapult/blob/master/docs
/histogram-set-json-format.md) |
| 45 data format, which stores full histograms and statistics about test |
| 46 results, as well as diagnostics with built-in UIs which help users dig in |
| 47 when something's wrong. |
| 48 * **Improve the sheriffing process.** We're working on burning down bugs in |
| 49 the sheriffing workflow, and also adding |
| 50 [milestone-level reports](https://docs.google.com/document/d/1MgTs2TBEHcKqgi
l_zqy72GlZr_74udyDKznTZixhrrI/edit) |
| 51 that give a high-level overview of how Chrome is performing from release to |
| 52 release, so that we can better focus our performance efforts. |
OLD | NEW |