Ha! I told myself so many times that I should take to Medium, but always postponed because I thought there are other community members much better positioned to do so, and doing it already very well. @veered, @sacha, and @diaconutheodor come to mind, but there are so many more. I guess I’m going to do it at some point.
You know, it just boils in me and almost spills over
< rant >
Meteor is beyond great, it helped us build an innovative product, a real business on top of it, and with a huge bonus in the form of, you know, hundreds of thousands of real people benefiting from it (e.g. the electrification projects).
The kind of network planning our data science partners deliver through our platform requires enormously expensive software. We make that a lot more accessible. And because it’s accessible, those at the receiving end of this knowledge, small electrical cooperatives, save millions of dollars. And because of those savings, many more villages get electricity.
You wouldn’t believe the impact…
We’re soon going to Haiti, and likely Africa and other places.
Then I see the FUD and downright stupid comments on HN, on this forum, and other places. I wonder how many of those, clearly n00bs, have ever built something really useful beyond “to do” apps with infinite scrolling pages, in need of the latest reactive renderer. Gee!
</ rant >
Regarding scaling, to be honest, Meteor and Mongo - the M&M - are small fish. We can serve 100 - 200 connected users from a local Mongo, and three instances of PM2 managed Meteor, all in one 4GB 2vCPU server, with plenty of room still left. Mongo can be rebuilt from data stored in Postgres easily if something happens. But we did not have a single issue because of Mongo. And we are sooo liberal with pub/sub, and believe it or not, actually composite publications most of them.
Our main concern is the tiling service, a Java beast running on a homebrew semi-elastic Tomcat cluster. Then we keep tons of spatial data in Postgres + PostGIS (talking tables with 20 - 30 million complex geometries sometimes). Plus a whole caching architecture. From there, we serve (tens of) millions of map features and data widgets to said users.
As @veered hints above, actually it takes a lot of traffic to really need to start talking seriously about scaling in the context of M&M. Until this day I really don’t understand what people are on about when complaining, particularly with redis-oplog available, or Grapher. It’s lack of coding skills, I tell you, not lack of options.