All models sent & in memory

[getguesstimate]

All models sent & in memory

[OAGr]

The great thing about loading in memory is that it makes everything really fast once things are loaded. The bad thing is that it takes forever to initially load. Progressive loading solutions can be tricky. Going to list data that needs to get loaded for different areas:

1. Initial Algolia query should return 30-50 responses, not 15. Loading more is annoying, there are lots now.

2. No server response should be necessary for page load. Just algolia. Right now we use Algolia to find the space IDs, then the in-memory store to show them, which is redundant as Algolia gives us the full entries.

3. A space should obviously load /space/ID.

We can keep on loading all the users for now, that data is much more minimal.

In the absolute worst case, we could simply just load the space on the space page, and then load the spaces index without the graphs elsewhere. This is probably a decent first step, should be pretty easy.

[OAGr]

This is my #1 priority at this point. Site is failing for a bunch of people, and may be down soon if we don't act quickly. Just a heads up.

[OAGr]

Thanks to the PR above, this seems fixed.

Was very interesting (for me). Basically, it seems like Rails auto cached responses. As /users and /spaces were called (this loaded everything in them), long-term memory use would rise. The quick rise after the dip at around 10pm was me doing a small bit of load testing on these urls. I also upgraded the database and reduced the amount it auto saves (in a good way, it used to try after each key stroke in numeric input, making the entire app slower).

Still some more work here, but I think this isn't a bottleneck any more (for now).