Non-reactive subscriptions or methods and scaling


I am trying to decide what would be the best approach for me in my usecase as I dont need direct mongo reactivity for now.

I am doing some fulltext search and few filters above indexed mongo database of ~20k documents .
For now it is using reactive subscriptions and no filters in deployed version. But locally I have some filtering.
Currently deployed on to test traffic and load for database inserts etc. I am working on all youtube livestream indexing too, but it is kinda tricky. For now all, and streams.

So 2 choices:

  • method calls to get whole list - easy to setup, but traffic can be quite higher as it needs to send whole list every time if I dont roll my own solution for deduplicating
  • low level non-reactive publish which would fetch same as method would do, but as they would be identified by _id’s DDP should filter out traffic for duplicates (yes there are some questions howto detect if I should call this.added or this.updated) but as there is no cursor observer, I would expect it can lower memory requirements.

And mini-mongo have nice querying instead of handling all in client side JS too.

What would you suggest from these, or some kind of mix.