Every time you launch an EC2 instance you make two decisions that affect everything else: which operating system it runs and how many resources it gets. Get the first one wrong and your application doesn’t work; get the second one wrong and you overpay. Here’s how I choose both.
Note: the video is in Spanish (Spanish audio only).
The image: your instance’s operating system
On AWS, operating systems come as AMIs (Amazon Machine Images), which is why you’ll hear the word “image” a lot.
The image depends on what you’ll run:
- If your service needs IIS to serve a website, you’ll most likely use Windows (Windows Server 2025 is selected by default, and you can pick another version from the dropdown).
- If you’ll use Nginx or Apache, a Linux distribution works.
I almost always use Amazon Linux: a distribution maintained and updated directly by Amazon Web Services. If you’re more comfortable with Ubuntu or Debian, they’re there too. And if your system isn’t on the list, click Browse more AMIs and search the AWS Marketplace (Fedora, for example).
Why I don’t use public images
There are also public images, which anyone can create and share. I don’t recommend them: precisely because anyone can upload them, they may contain vulnerabilities or backdoors. Only use images from verified providers.
Your own image: the Golden AMI
What I do recommend is creating your own AMI from an instance that already has everything you need installed. When you launch a new instance, you pick your image and it’s ready much faster. These are known as Golden AMIs.
Other image details
- Architecture: x86 or ARM, whichever works best for your application.
- AMI ID: the image’s unique identifier within your region. The same operating system has a different ID in another region; that’s how you make sure you always use exactly the same image.
- The image’s publication date.
- The user you connect with over SSH if it’s Linux. To create and manage the keys, see how to create SSH keys on AWS.
Instance type: start small
The instance type depends on your application’s workload. If you don’t know which one to pick, start with few resources and scale up. You can always go up a level when you need it, or down if you have too much. That’s why the cloud is called elastic.
In the video the default is a t2.micro: 1 virtual CPU and 1 GB of RAM (today the console may suggest another one, such as a t3.micro). There’s a smaller one, the nano, with 0.5 GB.
Sizes go like this: nano, micro, small, medium, large, 2xlarge, and so on. Each level doubles the resources, and it doubles the cost too.
Work out the cost before launching
The instance’s box shows its hourly price. For the month:
monthly cost ≈ hourly price × 744 hours
In the video I say a t2.micro comes to about 10 dollars a month; at North Virginia’s price it’s a bit less, around $8.60. Check your region’s price. And that’s not all you pay: storage is charged per GB reserved, and outbound traffic per GB sent.
Instance families
- General purpose: the T types (t2, t3, t4g).
- Compute optimized: like the c5n. The largest, the c5n.metal, has 72 virtual CPUs and 192 GB of RAM.
- There are also network, memory and storage optimized ones, and ones with GPUs.
Metal instances give you access to the physical hardware, for workloads that can’t be virtualized or need their own hypervisor.
One more thing: RDS, the relational database service, handles instance types and resources in a very similar way.
Choosing the right image and size lets you use the cloud’s elasticity: your application runs as it should and your budget stays under control.
Found it useful? Watch the full video on YouTube (in Spanish) and subscribe to @Andrevops for more videos like this one.




