[Osmf-talk] Tiles

[Steve Coast]

Paul this is great, thanks.

Let’s say you’re right. Then, in summary it costs no (or little) money, time or energy to run the tile servers.

Cluster A lists (A306) the following; "Action: Develop and deploy a vector tile stack for use with the OSM main website. This includes the software to import and update data and create vector tiles as well as the initial base style."

This sounds like a lot of unbudgeted work for little upside - as you say, the tile servers cost nothing (or not much) today. Therefore we could on the one hand start to budget all these hundreds of tasks and compare them to todays budget (which in the case of the tiles goes from 10% to some large percentage), organize them by desirability and if they will have any impact on anything we care about (say, completing the map). This would start to tie the many, many actions to reality. Or, we could just strike it entirely from the “cluster” and focus on a goal with an outcome.

Do you have numbers for all 26 “Cluster” A actions for their before and after budget costs, and any way to tie them to a measurable goal?

Best

Steve


On Jun 1, 2023, at 6:24 PM, Paul Norman <penorman at mac.com> wrote:

On 2023-06-01 1:31 p.m., Steve Coast wrote:
The OSMF is seeking massive new funding while at the same time operating a global minutely updated tileserver for the entire world to use for free. It's important to note that keeping this updated (for the entire world, for free) is itself painful and expensive compared to caching the tiles.

Several years ago, maintaining the tile service was painful due to trying to manage our own CDN over a large number of donated hosts. Since early 2021, this has not been the case and it is no longer a headache.

About 10% of the OWG's hosting and hardware budget<https://operations.osmfoundation.org/2022/12/31/plan.html> goes towards tiles. This is a mix of capital expense for servers; ongoing power, space, and network costs; and logging/monitoring costs. Each of those categories is very roughly a third. The actual spend is likely to be closer to 5% - tile-related costs are likely to come in under budget, but other savings might push the percentage back up.

All of the CDN costs and about half the render server costs are sponsored, thanks to Fastly, AWS, AARNet, University College London, OSUOSL, and Umeå University. If we were paying commercial prices, the CDN costs would be a significant portion of our budget, but we're getting the service for free.

Time costs are more difficult to tabulate, but because everything is well-scripted the tile service doesn't require much sysadmin time.

Today, bidenlaptopmedia.com, perhaps the United States most visited website in the last 6 hours, is now using OSMF managed tiles for free.

We publish top users of tiles at https://planet.openstreetmap.org/tile_logs/, but because the day hasn't ended yet, today's numbers aren't out yet. I had a look anyways.

Although their traffic is significant, the load from them is not. They are 54th in the US for the last 6 hours, based on backend tiles served, our standard measure of load. Odds are that users are viewing a small number of areas, so most traffic is being served by the CDN, which costs us no resources or sysadmin time.

Could we rethink this and instead limit tiles to editors, instead of seeking to expand an already free service which has plenty of commercial competition? That way, resources could be used on things that benefit OSM.

The vast majority of use from non-OSM sources is very small websites. To restrict access from these sites while allowing JOSM and editing-related sites like MapRoulette would require investments in access restricting technology which we investigated developing, but decided against because it wasn't worth it<https://github.com/openstreetmap/operations/issues/342#issuecomment-1138148939>. Developing this, administering it, and deciding what sites to allow would significantly boost the sysadmin time needed. It might also impact the willingness of sponsors to donate services, which could boost costs significantly and overwhelm any potential savings.

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/osmf-talk/attachments/20230602/b2254d1a/attachment.htm>