Thanks for that @filipenevola
I’m sure you are not using the same EC2 instance as Galaxy, Galaxy is running EC2 instances to host containers, you are probably running EC2 instances running your app directly on them, no?
Yes
On AWS I use that
Is there any way to get the same performance on galaxy ?
I had to lookup what an ECU was (it sounded familiar but I wasn’t sure). It turns out Amazon retired the term in 2014. This article runs through its history which is actually quite interesting. It’s clear this is a difficult thing to quantify. So does that mean the Galaxy ECU is still what it originally meant? In that article it says - the equivalent to an early-2006 1.7 GHz Xeon processor
On a t2 @sabativi each vCPU is equivilant to one thread in an up to 3.3 GHz Intel Scalable Processor. We started using the t3a which as a baseline has 2 threads in a 2.5 GHz AMD EPYC 7000 series processors. More details here
(FYI you aint gonna get anything better for $10 so my answer is only for when you get traffic really when you can afford $500+)
I’ve run galaxy. AWS and my own servers and out of all three my own servers out performed in both actual performance and cost. Buy your own dedicated cloud instance with a good datacenter near you, you will be able to call them direct for support and talk straight to the admin in the NoC and he will be able to go down to the rack and do whatever you need. I guess AWS is ok for real early cheap startups or people using credits from some program etc but it sucks because it’s so slow, it has outages, and everyone uses it so you get stuck in the queue.
The best is still to have your own servers and your own private lan between them so you can run mongo on it’s own box with no network overhead and have another machine doing whatever you need. You can load balance also very easily with this and AWS it’s a real PITA. Galaxy was easy to setup but expensive and very very slow.
Yes, but I really want to focus on code, I run a business with this app.
I am testing the double container and for now perf seems pretty much the same as before.
For now it is just two times more expensive than before ( paying beanstalk on AWS )
It makes sense at the start to try nickle and dime and get the best deal. Digital ocean provides the cheapest and fastest in my opinion although it isn’t much of a saving to AWS. One benefit is Digital Ocean will give you free hosting if you sign people up so if you have a blog you think you can get people to click your link on you can technically leverage that for free hosting, I did that for a couple years and never paid a dollar for hosting it because I was getting sign ups. It’s good while it works
It’s ok for Galaxy to be slow. It’s not ok for Galaxy to be expensive. Imho, it’s really not aligned anymore with the market.
We got our DB in London. Galaxy is in Dublin. We know it’s not ideal, it’s rather slow.
But for testing purposes I deployed a $15 Digital Ocean instance in Frankfurt (about same distance from London as Dublin) using Mup. That’s a 2 GB Memory, 1 vCPU server.
To match the speed of the DO instance with Galaxy, I need to upgrade Galaxy to Double containers priced around $160/month!
That’s 10x more expensive. I know you can’t compare Galaxy’s features with bare VPS. But still, that’s a serious incredible price difference to get the same performance.
Yeh and you really don’t need mup all you do is login and run the build - check the readme it provides when you build. If you did a digital ocean in their london point of presence, you can probably increase speed further. Galaxys features are free if you install nagios and a RRD tool like cacti and use Google analytics. So it’s a total nothing burger. You can get way better performance if you just RTFM and are prepared to run some commands.
Well, I don’t think it will motivate Tiny much to invest further in Meteor if you destroy the Galaxy business model… 
Besides, I think the biggest issue Galaxy had in the early days was that MDG tried to position it in the high-price market. Although Tiny changed that after the acquisition and even introduced a free-tier again (which is just awesome!), that perception still sticks to many Meteorists (including me). Renaming it to Meteor Cloud to get rid of that legacy was a good move.
We’re also on AWS now, and not likely to switch. Mostly, because we got used to it, and it just works. I can clearly see the benefits of an auto-scaling out-of-the-box service maintained by Meteor professionals, though. So there is definitely room for Galaxy. And we should keep in mind: that’s what motivates Tiny to keep up their good work in the end.
And thanks for the links to Nagios and Cacti. Wasn’t aware of these. One question here: Don’t these tools need to run on your own server as well, so you have to include this in your total price comparison?
Running your own server is about 6 times cheaper. Cacti and nagios don’t require much you can just run them on your dev server.
Honestly I see no point in AWS or galaxy beyond being a ten buck app / student.
You cannot auto-scale unless you run your own cloud. Running in bare-metal is just too slow (to scale) or too expensive (to keep emergency servers at bay if ever you need - and also pray the capacity is enough when you need to scale).
Scaling doesn’t work like that unless you are doing sports streaming or events.
If I need another server I call my admin and say add another server, it’s all virtualized but I’m not sharing with no one. That’s the difference. If I need another server it takes ten minutes, I run my scripts and it’s deployed. I add the ip to nginx config and the load balancer does it’s job. All in my own lan also. Costs about six to ten times less then AWS. AWS bandwidth is extremely overpriced. Leasing direct is far cheaper. Run the numbers and see.
That already negates your “I see no point in AWS or galaxy”
You have it back to front, If you running sports streams with peak viewers then you making bank already so why use it. When you get there you will get it. I’ve managed premier league, nba and boxing streams. AWS is a money pit that sends its ceo to the moon. Pay through the teeth if you want. Galaxy is a bad deal. If you don’t know Linux then you pay to not know how to RTFM, so it’s a good tax I guess. I manage 5 million daily active users and 45gbps bandwidth. If we ram AWS we’d be in debt.
Ad hominem. Classic. No point arguing further
Huh? I ain’t arguing with you, ain’t no beef man. I mean when you get to scale, then you will see the bills and see for yourself. Don’t get me wrong, AWS, Digital Ocean, Linnode etc is good for startups but when you start doing real numbers you will get crippled by the fee’s and that ain’t nothing personal. It’s just business. Amazon is a business, they are there to make money. They ain’t your friend bro
3.4x more expensive.
“ 1TB / month is $90 or $0.09 per GB
Again 3x higher cost than colocation, and 116x higher than Google Fiber, and if you have data coming in they charge you another $10 per month, making it $0.10 per GB.”
I am only trying to help. Don’t take it as a insult. I just mean when you get to a larger scale, be prepared for insane bills.
Thanks Vikr, I agree the bandwidth pricing is abusive. Every data center I’ve ever worked with in the past two decades, even before AWS existed was 365x24x7 with 99.9% uptime, these are standard marketing points.
Fault tolerance - Monit (why pay AWS to run a free gnugpl lib?)
Ddos? AWS doesn’t protect from ddos you have to run cloudflare and configure Nginx properly. I’ve been hired to do so for clients many times. The run is that AWS forces their dns route 56 I believe it’s called. Which is kinda janky.
Honestly you are more then welcome to use AWS. I really don’t care what you do. I’m just sharing my opinion that it’s overpriced and more suitable for startups, low / no budget student projects and windows users, in my experience it’s not viable to run a business at scale on Amazon, a direct lease with HP or Dell and paying the 95 percentile is far cheaper, plus you get dedicated hardware and ain’t in some pool with every other person, plus the entirety of the Amazon customer base. Each to their own. I won’t be using it. Needs to be far more competitive, not to mention Amazon’s policies for their own employees well-being… that’s the real turd in the punch bowl.
So here’s the numbers: on AWS 45gbps = $0.9 and that’s per second on AWS before server costs (Amazon charges $0.02 per gb so 45 * 0.02 = 0.9)
Per day I would be paying around $77k give or take.
Which means I save approx 2000% before any server costs.
