Hi,
Is there anyone here who understands EC2 enough to compare the pricing between hosting our own EC2 server and using Galaxy Developer Edition?
DGE is $0.035/GB/hour, or $13 per container per month.
The EC2 t2.micro costs just 0.013/GB/hour and each machine has 1Gb of memory available.
On the surface then EC2 seems a lot cheaper, and it’s not a spot instance.
What does DGE offer extra? Is it just the convenience of not having to set up the server with mup, or am I missing something else?
I don’t know what type of site you’re running with Meteor, but I have a web application (not a site), and from the start (when the code was lighter), I couldn’t get by without at least the t2.small. I’m now on t2.medium and things are smooth so far.
Does that suggest that several GDE containers would be needed to match that medium server?
Did you have problems due to the 512mb limit or due to CPU power?
I would assume you’ll need more than 1 container (unless you have a simple app / low volume of users).
You’re probably thinking that if galaxy is more expensive long term, then you are better off starting with something like ec2 or DO. I can understand that logic, but… getting started with galaxy is dead simple. meteor deploy simple. So why not just try it out initially and focus on solving other problems. Then in a few months if your hosting costs are too high, look to switch.
I don’t know if the galaxy costs are too high since I’m unsure what advantages I’m getting by using it.
It just seems to me that running mup deploy is just as simple, but perhaps there’s something I’m not seeing?
I suggest you just try it first then answer the question for yourself. Its pay per use. If you do go the galaxy route please post back again so everyone can learn from your experience.
I don’t think with mup + ec2 you can scale horizontally with a click of a button, but I could be wrong. You also don’t have to spend time figuring out how to deploy your app. Those two features, to me, would be the premium you’re paying for with Galaxy D.E.
Hmm…
If the price premium is purely so that you don’t have to learn to use mup, then I’d argue that it’s not worth it since mup is so simple. The time to learn it is negligible.
Scaling horizontally is a better argument to be sure. The meteor hacks cluster package makes this fairly simple, but it’s not the click of a button.
I have actually just tried Galaxy. For me the app ran very, very slowly compared to my existing test setup on an Ec2 micro, however the test isn’t fair since the Mongo database is in Europe and Galaxy is in the US.
So really is it fair to say that the biggest advantage is very simply horizontal scaling?
I haven’t used Galaxy yet (planned to try it once it was no longer invite only). Does it require you to manage the “server” - eg run commands like sudo apt-get upgrade or whatever to get the latest security updates etc or is all that managed for you? That’s the type of thing I was hoping to ditch when I move to Galaxy. I’m not a server admin and I don’t want to waste the time learning how to be one.
Edit: point being that I’d happily pay a premium for a managed server.
the GDE portal is pretty damn nice!
I also really rate the way they handle the container updates and how simple it all is to the end user.
I’m more than happy paying a premium to support MDG here.
I only wish that I could run more than 1 app for $13 a month.
During development my apps have 1 connected user and use a very small footprint. I guess I could spin up and down but I like to be able to show people and test on the run day to day so I leave it running.
You don’t do anything like that!
You just deploy to the server, and you’re done.
Galaxy builds a new container, migrates your connected users and kills the old one. It’s awesome!
Ah, now we’re getting to the crux of my question. I’m interested in what extra service I’m getting by using Galaxy compared to my own deployment.
If MDG keeps the servers updated with the latest security patches then that’s a big point. We’d have to do that manually on EC2 afaik.
You’ve brought up an interesting point about migrating users to the new container on a redeploy. I’d never thought of that. However, I can’t actually see how it’s possible since the newly running container would not have the merge box information from the previous one. Or does a redeploy simply result in all client data being resent? Does anyone here know what happens when redeploying using the docker-based version of mup?
From what I understand Galaxy now throws in SSL certification for you domain for free. That’s slightly involved to get setup using MUPX and also extra $$$. Like $14 - $25 per year depending on what kind of certificate you buy.
It’s no problem at all to use Let’s encrypt certificates with mupx. I’m doing it, and it works fine. Only problem is that you have to extend the certificate duration every 3 months. But there are PRs to automate this.
HELP: Does anyone know how to get a mupx instance working, where you have to have read/write access to the file system OUTSIDE of Meteor proper? I have a bunch of templates and temp files I use and don’t want to copy each one every time I deploy. This is what’s keeping me from mupx (aka docker).