Benchmark client side rendering performance
Closed, ResolvedPublic3 Estimated Story Points
Actions

Assigned To

Authored By

	Jgiannelos
	Dec 15 2020, 3:53 PM

Description

Context

Part of this iteration is to evaluate if client side rendering has an improvement in the experience of using maps in wikipedia projects.

Conclusion

The data report can be viewed in this spreadsheet

All frameworks and libraries were tested with a simple VueJS App. No efforts to improve performance were made in the app per se, but all the tested options had the same baselines for lib loading and vector-tile server source (except for tangram)

Acceptance criteria

Report for frontend rendering timings compared to consuming raster images

Details

	Subject	Repo	Branch	Lines +/-
	client-rendering: benchmark MVT and raster tiles	maps/spikes	master	+15 K -0

Customize query in gerrit

Related Objects
Search...

		Status	Subtype	Assigned	Task
		Open		None	T270188 [EPIC] Client-side vector tile rendering
		Resolved		MSantos	T270190 Benchmark client side rendering performance

Event Timeline

Jgiannelos created this task.Dec 15 2020, 3:53 PM

Jgiannelos moved this task from Needs triage to Backlog on the Product-Infrastructure-Team-Backlog-Deprecated board.Dec 16 2020, 11:38 AM

Jgiannelos moved this task from Backlog to Upcoming on the Product-Infrastructure-Team-Backlog-Deprecated board.

Jgiannelos triaged this task as Medium priority.Dec 16 2020, 12:32 PM

MSantos moved this task from Upcoming to Kanban on the Product-Infrastructure-Team-Backlog-Deprecated board.Jan 7 2021, 4:35 PM

MSantos edited projects, added Product-Infrastructure-Team-Backlog-Deprecated (Kanban); removed Product-Infrastructure-Team-Backlog-Deprecated.

MSantos set the point value for this task to 3.

MSantos claimed this task.Jan 11 2021, 9:59 AM

MSantos moved this task from To Do to Doing on the Product-Infrastructure-Team-Backlog-Deprecated (Kanban) board.

MSantos updated the task description. (Show Details)Jan 13 2021, 4:00 PM

MSantos moved this task from Doing to Code Review on the Product-Infrastructure-Team-Backlog-Deprecated (Kanban) board.

It looks very interesting. Overall it seems like all the metrics between engines are similar except of the total blocking time.
I am wondering how the numbers would change if the benchmark setup is not just a main div in the DOM but also includes other elements where a map is nested.

MSantos moved this task from Code Review to Sign off on the Product-Infrastructure-Team-Backlog-Deprecated (Kanban) board.Feb 2 2021, 12:13 PM

MSantos closed this task as Resolved.Feb 16 2021, 10:22 AM

@MSantos Just a small note regarding map loading performance. We currently use a single hostname for serving up tiles. Especially when multiple tiles are in the map (fullscreen), the browser tends to parallellize no more than 5 download requests for png image per host at a time. In my experience this is actually the biggest delay when loading any sort of larger map view that includes multiple tiles (fullscreen). This is also why on the old toolserver we used to have tiles.a.toolserver.og, tiles.b.toolserver.org and tiles.c.toolserver.org and why the leaflet style url specification allows for multiple domains.

Not sure if that needs to be taken into account, just wanted to share in case it might affect your metrics.

Change 664835 had a related patch set uploaded (by MSantos; owner: MSantos):
[maps/spikes@master] client-rendering: benchmark MVT and raster tiles

https://gerrit.wikimedia.org/r/664835

gerritbot added a project: Patch-For-Review.Feb 17 2021, 3:10 PM

Change 664835 merged by MSantos:
[maps/spikes@master] client-rendering: benchmark MVT and raster tiles