Asia pacific(Mumbai) p2 instance

When requesting for the p2 instance at aws, can i select Asia pacific(Mumbai) region since I an Indian?
is the ami available for that region?

I would advise against that. AWS Mumbai pricing is on the higher side. Check the pricing for each region here:


WOAH! Thanks a lot. You just saved me :slight_smile:

1 Like

So which one to choose?

Just check the price here - and go with the region that’s closest to you and reasonably priced. I use US West (Oregon) because it’s closest to AWS :smiley:

Also you have to see if the AMI we will be using is in the region you want. Most likely Jeremy will create the AMI in some US region. I am sure there are ways to copy and move it to the region you want, but I think the prices will be similar and US region prices might be cheaper than others.


For the Asia region it seems Mumbai has lower costs relatively…

I feel US West Oregan is a better choice with relatively lower price and almost always gets the latest hardware extensions early than the other regions. If you are thinking about near by regions for better connectivity then don’t bother. It won’t make much of difference.

1 Like

The decision to choose among different AWS regions depends on multiple factors, with the primary ones being:

  1. Cost
  2. Latency
  3. Security

Because we are not building services that need to be accessed by a lot of people, we should be optimizing for cost.

I found cheapest amazonec2 to be quite useful.
At the time of writing, it suggests us-east-2 for p2.xlarge spot instance as it costs 0.1301/hour.

I believe Fast AI has not released AMI instances that are configured with the Deep Learning stack required for running the notebooks in Asia-Pacific(Mumbai) region.
ie you cannot get a p2 instance that is similar to the one used by FastAI in Mumbai region…
I am saying this from the info I got from the scripts( as it supports only 3 regions ( us-east, us-west, eu-west)

If you choose to spawn a p2 instance from Mumbai. I feel you need to download anaconda , keras & tensorflow by yourself .

I am using the eu-west AMI, I don’t see much latency. Its practically usable. I just finished Lesson 1.

Have you guys considered Or using aws spot instance?