-
Notifications
You must be signed in to change notification settings - Fork 1
Charter 2: Explore backend performance to discover surprises
mwalker-scottlogic edited this page Jun 4, 2024
·
6 revisions
Explore backend performance to discover surprises
- Mike Walker-Rose
- 28/05/24, 04/06/24
- 45 mins
- This charter is focused on improvements in performance, which can be measured using the logging of the system. It used to take 10 minutes to perform ETL, this should be faster while having not broken the system.
- N/A
- running after CAMS update
- running after CAMS update (00:02:17.985)
- running again after CAMS update (before next) (Result: 00:00:46.294)
- And again... (Result: 00:00:46.25)
- Delete database data & re-run (Result: 00:00:55.586)
- Delete grib files and re-run (Result: 00:01:24:509)
- Delete grib files & database data and re-run (Result: 00:02:03.754)
- Re-run forecast (Result: 00:00:46.201)
- Delete grib files and re-run (Result: 00:02:05.325)
- Delete grib files & database data and re-run (Result: 00:02:04.835)
- Delete database data & re-run (Result: 00:00:59.114)
- Delete grib files and re-run (Result: 00:01:32.182)
Version | Time |
---|---|
Initial run after new model base date | circa 2:17 |
subsequent runs with existing gribs and database data | circa 46s |
running the process using the same grib files, but deleting the db | circa 55-60s |
running the process using the same db data but deleting the grib files | circa 1:24 - 2:05 * |
running the process after deleting db data and grib files | circa 2:04 |
(*however 2 of 3 test runs were in range of 1:24 - 1:32, possible human error in outlier?)
- running after CAMS update (00:01:14.499)
- re-running (00:00:18:956)
- re-running (00:00:19:893)
- Delete database data & re-run (00:00:37.895)
- Delete database data & re-run (00:00:32.17)
- Delete grib files & database data & re-run (00:01:16.459)
- Delete grib files & database data & re-run (00:00:58.869)
- Delete grib files & re-run (00:00:55.886)
- Delete grib files & re-run (00:00:36.212)
Version | Time |
---|---|
Initial run after new model base date | circa 1:14 |
Subsequent runs with existing gribs and database data | 19 - 20s |
Delete only database data & re-run | 32 - 37s |
Delete only grib files & re-run | 36 - 55s |
running the process after deleting db data and grib files | 58s - 1:16 |
Getting Started and Overview
- Product Description
- Roles and Responsibilities
- User Roles and Goals
- Architectural Design
- Iterations
- Decision Records
- Summary Page Explanation
- Deployment Guide
- Working Practices
- Q&A
Investigations and Notebooks
- CAMs Schema
- Exploratory Notebooks
- Forecast ETL Process
- In Situ air pollution data sources
- Notebook: OpenAQ data overview
- Notebook: Unit conversion
- Data Archive Considerations
Manual Test Charters
- Charter 1 (Comparing ECMWF forecast to database values)
- Charter 2 (Backend performance)
- Charter 3 (Forecast range implementation)
- Charter 4 (In situ bad data)
- Charter 5 (Filtering ppm units)
- Charter 7 (Forecast API input validation)
- Charter 8 (Forecast API database sizes)
- Charter 9 (Measurements summary API input validation)
- Charter 10 (Seeding bad data)
- Charter 11 ()Measurements API input validation
- Charter 12 (Validating echart plot accuracy)
- Charter 13 (Explore UI after data outage)
- Charter 14 (City page address)
- Charter 15 (BugFix diff 0 calculation)
- Charter 16 (City page chart data mocking)
- Charter 17 (Summary table logic)
- Charter 18 (AQI chart colour banding)
- Charter 19 (City page screen sizes)
- Charter 20 (Date picker)
- Charter 21 (Graph consistency)
- Charter 22 (High measurement values)
- Charter 23 (ppm -> µg m³)
- Charter 24 (Textures API input validation)
- Charter 25 (Graph line colours)
- Charter 26 (Fill in gaps in forecast)
- Charter 27 (Graph behaviour with mock data)
- Charter 28 (Summary table accuracy)
- Re‐execute: Charter 28
- Charter 29 (Fill in gaps in situ)
- Charter 30 (Forecast window)
- Charter 31 (UI screen sizes)