Amazon enters the "consumer" level VPS market


#1

For anyone who hasn’t seen this yet:

https://amazonlightsail.com/

Competing with the likes of DO, apparently. This is a great option for those who find AWS either too complex or don’t like the uncertain pricing.


#2

That’s great news. More competition =)

The only problem I see is that it’s still VMs. I made the switch from VMs to Azure Apps and there’s no way I’m going back to managing VMs.

or don’t like the uncertain pricing

Which services have “uncertain pricing”?


#3

Just out of interest, what’s uncertain about their prices?

As I see it it’s just an easier management UI in front of their t2 EC2 instances. It’s slightly more expensive than diy with EC2 in that regard (between 5 and 10%).


#4

Mh, I think it is very difficult to compare VPS systems, because a lot providers don’t tell the details in their offers. What is 1 core processor? Which clock rate does it have? If I take a look to OVH, they seem to be much cheaper and also running their VPS systems within a cloud. The same for Hetzner (also running within a cloud due to their FAQ).


#5

Well as they’re running T2 instances there is a lot of info about those (do quick google search and you’ll find a lot of details and benchmarks). They use a burstable model with a certain amount of cpu credits per hour. I think the base rate (if you don’t have any credits left) is around 20% of the full capability of one xeon core, which is burstable to 100%.


#6

Uff…okay, I have to say I’ve never used any AWS services. So I don’t get “dedicated” resources if I choose the $80 offer (2 core processor) - the CPU power of those depends on the credits?

I’m just trying to compare their offer with others - just had the same problem with DO a while ago because I didn’t know what I really get.


#7

Yeah the T2 instances are built on the burst principle which most of the time is actually really effective. Unused credits are stored for 24 hours so it handles traffic variance well in most cases. If you want continuous full use of a cpu you’d need to look at their other instance types like M3 or M4. A large M4 instance for example will cost you almost exactly $80/month but gives full access to 2 cpu cores, compared to a T2 large instance which costs around $70.

The single best way to determine what you’re actually getting is to run benchmarks, most ones that you find are not at a sustained rate or are manipulated towards one or the other.

I use aws exclusively, if you make use of their spot instances you can really get cheap deals, the same M4 large instance would only cost about 12$/month when using current spot prices for example. Combine that with ECS (docker platform) and you get a very scaleable, manageable and cheap infrastructure for your servers. For beginners or small products this might not be suitable though, so in that case either DigitalOcean or AWS might be a good fit.


#8

What I meant by “uncertain prices”:

Not that the prices aren’t clear, but the consumption of the various aspects of EC2 (EBS storage, bandwidth, etc) aren’t fixed. Some people might be uneasy about this and are more comfortable with a $5/mo fixed price.


#9

@manuel BTW, if Azure is not VMs, what is it? Surely they’re not giving each user their own real machine.


#10

Besides VMs Azure has what they call “App Service” which is a PAAS (like Heroku). So you pay for the resources you want to use (say 1 core, 2GB ram, 10GB hd, etc. plus how many instances you want to run), and then you can deploy as many apps as you want and they all will use the resources you’ve allocated to that group.

You can create as many groups as you want with different resources allocated to them, and every app deployed to a group will share the resources with others in the group. Of course nothing stops you from deploying a single app to a group so it has all the resources available to it.

Deploying an app is as simple as telling Azure where your repository is and Azure will do the rest. Deploying new versions is as easy as pushing to master.

It’s more expensive but a hell of a lot simpler than managing VMs.


#11

That still sounds like a virtualized container/VM. By VM do you mean a straight-up plain VPS?

Lots of people seem to like Azure for Meteor deployment. I’ll have to check it out next time I want to launch something and don’t want to invest in Galaxy. Thanks!


#12

By virtual machine I mean just that (a VPS). I have deployed docker containers to my own VMs but I haven’t dealt with a service that allows me to simply upload a docker container directly (no VMs involved).

Azure Apps work more like Galaxy. You give it your app and then set up the resources you want it to have available to it. I like it better because I can deploy a bunch of little apps to the same group without paying additionally for each.


#13

That sounds awesome! I’ll give it a run when I get a chance.


#14

That’s exactly the same as AWS ECS, you build a cluster of vm’s (by defining an instance type and the amount, same as defining resources in essence) on which you can deploy as many docker images/instances as you want. They run on normal VPS’s that run a default linux image so they’re still VM’s, you just don’t manage the separate servers directly and have some automation tools available like autoscaling and load balancing.

If you have a somewhat larger app it’s the way to go imo, either Azure apps, ECS or google container service are good fits in that case.


#15

I tried Azure just briefly. I jumped ship also pretty quickly. :wink: I didn’t have anything against the service, but it felt like Microsoft’s version of AWS, and after investing so much time learning AWS (which is daunting, as many people know), I can’t really afford the time learning another AWS-like platform. I can just as easily launch an EC2, get a super low fee by purchasing a reserved instance, and use mup + nginx to deploy multiple apps to a single server.