Google Colab vs. Microsoft Azure notebook

Although I knew this service for a while, I just recently put attention on 2 “serverless” notebook services on the cloud: Google Colab and Microsoft Azure Notebooks.

Here are my short reviews.

<<Google colab>>
1. only support python (currently 3.6.7 and 2.7.15). you can build the packages through pip directly from the notebook. no way to build an isolated environment such as conda.
2. allocate big RAM (12 GB) and enough disks (50 GB)
3. GPU is available by default. you can use it for anything except crypto mining and long-term usage!
4. your codes will be saved in google drive and could be edited collaboratively (by sharing)
5. integrated with cloud storage such as Amazon s3. just mention the file in s3 bucket (s3://[path]).
6. have the capability to use local hardware, why? maintain the collaboration process and extend the hardware (if you have a better one)
7. integrated with code development in a team such as github

<<Azure notebook>>
1. multiple kernels supported: python, R, F#
2. small RAM (only 4GB)
3. no available GPU to use, you should deploy your own VM to have GPU
4. retrieving s3 data takes so much time
5. not sure how to use local hardware
6. integrated with github

Leave a Reply

Your email address will not be published. Required fields are marked *