Versions Compared

Key

  • This line was added.
  • This line was removed.
  • Formatting was changed.

Columbia University has arrangements with a number of cloud providers, below is a list of these with the other resources available to academics and the general public which should which will help get you started in the cloud. While Research Computing does not directly support the cloud, we can help you get started, if you have any questions that aren't answered in our wiki or documentation drop us a lineWe can assist you with guidance and recommendations that are appropriate for your specific use case. Please email us at rc@zi.columbia.edu to get started or schedule a free consultation.

Amazon Web Services (AWS)

Amazon Web Services is the largest of the cloud providers for good reasonworld’s largest and most broadly adopted cloud platform, offering over 200 fully featured services from data centers globally. They pioneered the modern cloud paradigm and have continued to launch new services revolutionizing the potential for leveraging on-demand resources on the fly.

Getting an AWS account

Columbia has Business Associate Agreement (BAA) with AWS, so if you are a faculty member or administrator who uses AWS for Columbia purposes, your AWS account is required to go through the CUITmust be requested through CUIT, which manages the creation of new Columbia AWS accounts. Please email us at rc@zi.columbia.edu for assistance getting started or you may directly request your account here: CUIT Amazon Web Services. Student and personal AWS accounts are not covered under the enterprise agreement.You  You will need to provide a chartstring payment Chartstring to request an account and there is a good deal of documentation to read, including what is and is not permitted to use a Columbia AWS to doyou will be billed monthly.

Columbia AWS resources

...

Internet2 - Columbia is a member of the Internet2 consortium and researchers are eligible for a data egress discount by applying for it.

...

  • courses for free which has a number of training classes for general cloud practices and AWS in particular.
  • AWS specialists - Columbia has

...

  • access to cloud architects and other AWS specialists as part of our relationship with Amazon.

...

Additional AWS resources

Google Cloud Platform (GCP)

...

GCP accounts can be linked to CUIT through the process listed on their Web site CUIT's Google Cloud Platform website, which lists the advantages of linking your project to Columbia, including the ability to pay through the Columbia accounting system, and links to the uni Columbia UNI authentication system.

...

Google Cloud Platform resources

...

Other Google resources

Lionmail Drive comes as part of all Lionmail email accounts and Columbia's agreement allows for unlimited storage, however there are, undocumented, transfer throttling limits in place, which limit its usefulness.

Colab is a free service which offers access to Jupyter notebooks with attached GPUs, and can leverage Lionmail Drive for storage. Colab 101.

...

titleGoogle Workspace (formerly G Suite) vs GCP

...

Please note that GPU resources available for free on Colab are limited. To reserve additional computational resources such as faster GPUs, purchasing compute units, or access to more memory, you may need to upgrade to paid subscription or purchase compute units on-demand as shown here: https://colab.research.google.com/signup

Microsoft Azure

Columbia has a Business Associate Agreement (BAA) with Azure via the CUMC CUIMC IT, however at this time there is no special Columbia process or documentation for setting up Azure accounts.

...

Additional cloud resources

DigitalOcean

Lambda Labs

Click the chart to view larger imageImage Added

For researchers looking for cost-effective GPU resources, Lambda Cloud provides powerful on-demand GPUs at highly competitive rates compared to other major cloud providers like Azure, AWS, or GCP. These GPU instances are pre-configured for AI / machine learning.

While Lambda Cloud currently does not support pausing or suspending instances, they are actively working on this feature. In the meantime, Lambda Labs recommends using Docker containers and persistent storage options to save the state of the system, making it easier to resume workflows as needed. For more information, see their documentation on instance management and persistent storage. Click the chart to the left to view hourly pricing.

Using the Cloud/AWS at Zuckerman 10/16/

...

2020 presentation slides (requires Lionmail account)