Scientific Python in the cloud

Using Python for data analysis in the cloud

Several services allow you to use Python without installing any software on your own machine. These services provide servers on which the scientific Python stack is installed. Users connect to these services through a web interface, so all you need to make use of them is a computer with a web browser. Many people see this approach as the future of scientific computing, but the current generation of cloud computing services for scientific Python has some limitations. Nevertheless, it is usable, especially for training and learning.

• [https://cloud.sagemath.com](Sage math cloud) -- this service provides both the IPython notebook and their own notebook format. In addition to Python, they provide R, C, Fortran, and several other tools. Currently all accounts are free. We have seen some service degradation in mid-day.

• https://www.wakari.io -- a commercial service provided by Continuum Analytics. They provide free accounts, but these are often unusable during busy times of day, and their free system is too slow for most real analyses. However it is quite useful for learning, teaching, and training. The system exclusively uses the IPython notebook, and does not support languages other than Python. It is fairly easy to configure your own environment (e.g. loading non-standard Python libraries).

• The SCS servers (scs.itd.umich.edu) -- a U-M internal service, the "old school cloud" is a set of Linux servers. To use this system, you need to learn how to use ssh to connect to a remote machine, and use terminal tools like text editors (Emacs or Vim). These machines currently lack an updated Python distribution with all the libraries discussed here, but it is possible to install everything in "user space".