Which path will you follow for you Meteor apps and why?
Having done both, I feel there are a lot less headaches in terms of set-up and maintenance with Compose. And you get a nice, unified web interface to interact with your data, do backups, view logs, etc. I feel safer as well … it’s those guys job to know what they’re doing with mongo and to optimize things and ensure there are failovers.
With my DIY deployment, I somehow feel much more susceptible to disaster …
So, at least in the medium term, I’ll be sticking with PaaS db services like compose. At $18 a month for the basic plan with oplog access, it’s not killing me financially either.
We are doing the both.
Managing a Mongo installation is not simple.
So, we use Compose for the Kadira UI.
But, we keep a lot of data in MongoDB as well for metrics.
we can’t simply use Compose for that since it’s too much costly for us.
For that we host it our selves and monitor it regularly.
We could’ve done the same for the Kadira UI. But, moving that to Compose reduces some of our responsibilities.
Scaling options and best practices?
I think compose and the likes are good for mid-size requirements.
For small things, getting a mongo replica set up and ready takes less than 5 minutes and $5/month on digitalocean.
For growing projects, you’ll most likely have a devops guy whose job is to manage your infrastructure.
But for those projects that lie in between, I believe there’s the area you’ll benefit most from paas mongodb solutions in terms of price/performance.
I think Kadira is a particular app, because the analytics data expire after a week.
It seems more a taste of responsibilities, backups policy and nice UIs
Probably, in the long time, an infrastructure you completly own is cheaper if you plan to reuse it for several projects.
I’m using Compose and it’s just amazing, saving us a lot of time and letting us focus on the product. Plus their support is really good and they know their stuff.
As for large data sets or bigger projects, we have this and looking at the numbers for cost. Compose seems to still be a better solution. Yes it’s a bit more expensive but still cheaper then trying to maintain everything on our end.
I am also facing this decision. I think there are latency issues if the the sass server (compose…) is not in the same data center as your meteor server.
Also I am concerned about security, because compose doesn’t offer ssl / ssh-tunneling, or am I wrong?
Edit: No ssl, but ssh-tunneling is possible (but probably more latency)