UserStatus.connections memory usage

[rclai]

Will there be a lot of memory consumption if the UserStatus.connections collection has like.. millions of user connections in it?

[mizzao]

No, it's in the database, but there will be memory usage based on publications because of the merge box.

Although, having a million connections probably requires you to have multiple node processes, which currently isn't supported on this package: #21.

[rclai]

If that collection is in the database why does it have a collection null option?

That was a very informative but also somewhat discouraging article. I love Meteor but I hope MDG figures out how to solve the merge box issue.

[mizzao]

Oh @rclai, you're right. It's not in the database. I was thinking of Meteor.users for some reason.

Anyway, other things will cause your app to crash way before you reach a million users, not this :)

[rclai]

Okay cool. I was wondering what your feedback is about this article and whether it has any relevance to this merge box issue?

https://kadira.io/academy/optimize-memory-usage/

[mizzao]

That talks about sharing observers (in the facts package this is known as observe multiplexers), which can cut down memory usage for identical queries between clients. But that's just for managing queries; each merge box is a cache of clients' data which still has to be maintained.

It also mentions using a method instead of a publication, which will help with this issue by saving memory for large publications that are updated infrequently. But it will be more work to implement, so avoid premature optimization :)