Hey everyone! I'm new to AWS and I'm looking to get some advice on a project I want to deploy. I have an image processing project in Python that I've developed locally, but it maxes out my 8-core i7 and uses around 40GB of RAM, which makes typical hosting services unsuitable for me. Instead, I'm considering using EC2, but I haven't set it up before, so I have a few questions: 1. Is it really as easy to set up EC2 as it seems? Once I create an instance, can I use it like a regular computer with a desktop interface? I saw guides on running a web server with Python, which is my focus. 2. If I need more RAM or better hardware mid-development, will I have to reinstall Linux drivers if I change or add components like a GPU? 3. What should I keep in mind when choosing hardware? I need about 64GB of RAM, a solid CPU, maybe a GPU, and around 100GB of storage. I was looking into c6g.8xlarge or c6gd.8xlarge but could use some advice on hardware. 4. Lastly, how does billing work? I assume I only get charged for the time the server is on and not when it's stopped, right? And if c6gd.8xlarge costs about $1.30 per hour, is that accurate? I want to know if I can afford something a bit pricier, as my project is somewhat resource-intensive. Thanks!
3 Answers
Good choice going with EC2! Just a heads up, if you're considering using a GPU instance, new AWS accounts start with zero quota for those, so be ready to submit a quota increase request if you plan to use one. Also, check out instances.vantage.sh for a better overview of instance types and pricing—it's super helpful! Just remember, AWS bills you for both what’s running and what’s provisioned. So if your EC2 server is stopped, you won’t be charged for the instance itself but you will pay for attached storage and any public IPs. And don’t forget to set up cost alerts early to avoid nasty surprises!
You’re on the right track for a proof of concept (POC) with EC2. You can definitely pick an Ubuntu AMI, fire it up, and SSH in. Using a desktop interface is possible, but most people just use SSH for simplicity. Upgrading your instance size is straightforward—just stop it, select a new size, and reboot. For GPUs, you'll need to choose a specific instance family that supports them right from the start, as you can’t add one later. Be cautious with the c6g family since it's ARM-based, and not all Python packages work well there. I recommend looking into m6i or c6i instances for x86 compatibility. If you need GPU power, check out g4dn or g5 instances for your needs.
Thanks for the tips! I didn’t realize the architecture mattered so much. For GPU options, which instance would you recommend for deep learning frameworks like TensorFlow or PyTorch? And is my cost calculation correct?
Before diving into EC2, think about how long those image processing jobs will take. Sometimes, Lambda functions can be a more cost-effective solution for Python workloads. They handle the backend overhead for you, meaning less management hassle! Plus, they’re designed to save you money if you can fit your processing within their constraints.
Can you set a budget limit on services? Like, if my costs hit $50, can I have it stop everything and notify me?