Open rafaelspring opened 3 years ago
Hi rafaelspring,
I don't know how to do this. But I think http://atis.cloud/de Atis.cloud have managed to do this. Since we discovered this solution, we have stopped developing an internal service. Because as our boss told us, the cost is so low that it's a better deal than having in-house developers. You can also have it white label to rebrand it with your company.
Not gonna lie, I'm kinda surprised they're not a massive donor to potree, all things considered. If they have 300 clients subscribed, and the minimum for their advanced (not custom) level is about €100 each month, that's over a quarter million euros a year. 10K seems kind of trivial at that point. But maybe the funding page is outdated. 😕
Hi rafaelspring , I am working on it on my side, if I manage to do it, I'll let you know. Dockerized was an easy one but lambda I'm not quite sure it is a good idea, but i keep on !
@jbijoux Thanks! Where's the best place to deploy this as a Docker container?
Hi rafaelspring , I am working on it on my side, if I manage to do it, I'll let you know. Dockerized was an easy one but lambda I'm not quite sure it is a good idea, but i keep on !
@jbijoux Hey Julien! We tried to dockerize the Potree Converter but faced some difficulties. Could you please share how you've done it? That would be truly helpful! Thanks in advance!
Is there any experience with deploying PotreeConverter as a cloud (serverless) function?
The problem we're running into is that the memory footprint for point cloud conversion can be very high, sometimes many gigabytes. However most serverless providers (Google, Azure, AWS Lambda) have pretty tight memory limits of 2GB to 4GB maximum.
What are ways to solve this? We'd like to avoid the overhead of engineering the scaling / instance provisioning etc. ourselves.