AWS hates it: One weird trick to get the most mileage out of your money / credits

Being frugal for the sake of being frugal is not very useful. Your time has value and you could spend it on learning / relaxing vs investing it into saving a little bit of money here and there. Meaning, sometimes spending more means you get more value, sometimes it doesn’t.

Having said that, we all come from different walks of life and maybe for someone this advice will be genuinely helpful. I know I have resolved to using this myself on a couple of occasions myself for various reasons.

Here it is - for 2 cents you can get an hour of usage of a m3.medium spot instance! You also need to pay for the EBS storage but according to the docs it comes with a 4 GB ssd, which will set you back 40 cents per month of usage! That is 40 / (24 * 7 * 4.5) cents per hour.

Why would you want to take advantage of this? For one thing, it is quite neat to use a fully functional instance at such an attractive price. But, more importantly, chances are there will be many, many uses where this instance will be perfect for this course.

Do you know git? If you are planning on using course materials and do any development yourself, this is a tool that you probably should learn. You can do it easily on this instance.

What about python? Do you understand how data gets moved around for the training and the commands in the notebook? Another perfect use case.

Do you know pytorch? I am seeing pytorch for the first time in my life! To familiarize myself with the API, do a couple of tutorials, I don’t need a GPU - not sure if the smallest m3 instance available is the best choice here but who knows. If not, I can spend the extra 0.5 cent per hour and get the m3.xlarge. At 2 vCPUs and 7.5 GB of RAM it is a force to be reckoned with!

And yes, you could probably do a lot of this on your local machine, but doing this on the instance gets you in the practice of using a server in the cloud and you get to learn all the bit associated with this! How to create the instance, how ssh works, what can you do to edit code, how to install stuff, you might even be interested in learning tmux, etc. All the really good things especially if for some strange reasons you are stuck with a local Windows machine :wink: (still you probably should be installing a VM or a linux subsystem ASAP but that is a biased opinion of a random person on the Internet so take it with a grain of salt :slight_smile: )

Sorry if my post seems a little bit light hearted but it is Tuesday morning, I am super happy that this course is happening and grateful that I am part of it, so guess it is okay if I let some of the good mood seep into the post! A happy Tuesday to you all and maybe some of the above information you will find useful! :slight_smile:


Hi @radek i am following all your post, really complete information, thank you very much.


Hi @radek … It appears to be a very good advice . But being completely new to AWS and other stuffs I didn’t follow stuffs you just described over there . If you could give a little extra insight it would be quite helpful. Also, do notify @jeremy so that he could show it practically in his next lecture (as we shall be actually start using AWS from there on ) if it benefits everyone…

Hi @Vishucyrus - I can absolutely see how AWS can be tricky. They seem to have a thing for using a lot of jargon that is next to impossible to decipher outside of corporate IT and also they tend to come up with their own names for things.

I think a great starting point would be seeking out AWS tutorials to get your feet wet and see how things work. Does anyone have any tutorials they might suggest? Or any good AWS intro videos?

I try not to focus too much on AWS besides the basics so that I can get things done. I follow the set up as I outline in this medium post which is to a large extent based on the set up from part1 v1 of this course, with slight tweaks to automate things. There are many similar write ups, probably the most popular one being this one by a former student of this course, @slavivanov :).

Once you feel ready I would encourage you to pick one of these tutorials and follow the steps even if you do not fully understand what is happening. No worries, you will learn things as you go. This is a key insight I took away from @jeremy and @rachel and have to keep reminding myself of that and can fully relate to the comments @jeremy shared in the lecture yesterday from former students saying that they wished they focused less on theory and more on actual coding / doing.

Anyhow, I digress. See if any of the materials I list above might be of interest to you and also let’s see if anyone has any AWS related howtos or introductory material they particularly enjoyed and wouldn’t mind sharing :slight_smile: Maybe even starting a separate thread for that would be of use? If anyone feels that would be a good idea, please go ahead and start it!


I recommend A Cloud Guru (best AWS online tutorial I have found):
I started with this one:
It is 22 hours, but you can go through different topic sections to get specific info, though it’s good to begin from the start.


+1, i did this course too - it’s quite thorough… Richard Jones on O’Reilly Safari is excellent but you need to pay a subscription fee

1 Like

For $0.04/hour you can use Crestle with GPU disabled - benefit of that is that it’s simpler, and you can switch between GPU on and off without moving your data around. But it’s a little more expensive that raw spot instance.


yes but i have 500USD for AWS, and I want to use it as best as possible, Crestle was good for the first class, but I see that it doesn’t do very well with many users, and this should be reviewed in the long run.

1 Like