Costing the cloud in proposals

Introduction

The purpose of this page is to provide guidelines for proposals that incorporate cloud computing, specifically towards helping you (the proposal author) estimate costs for your project.

Warnings

Overview

  • Review your status and options with regard to cloud research credit grants.
  • Verify with the funding agency that cloud computing will be a welcome component of your proposal.
  • Remember that the authors of this website have office hours and are happy to consult with you.
  • Create a cost plan per year
    • Estimate the compute power per machine and number of VMs you will want to secure
    • Estimate the cost of associated memory
    • Estimate the cost of separate storage
    • Estimate your egress cost (cost of retrieving data from the cloud)
    • Estimate the cost of other services you intend to use
  • As necessary: Provide justification of your choice of cloud
    • You can compare to alternative cloud providers
    • You can refer to the advantages of using the cloud over purchasing and maintaining equipment
    • Cite case studies presented here at cloudmaven.org
    • Remember that the case for cloud includes strengths:
      • Computing scale
      • Data security
      • Advanced tools and services
      • Similarity of compute environment to local machines
      • Rapidly growing user base and commensurate open resources from which to build your solutions

Cloud Research Credits

Generous research credit grants (circa 2017) are commonly provided by cloud vendors, specifically Microsoft Azure and AWS (and GCP grants may materialize as well). These grants can have a value as high as $20,000 and a typical duration of 12 months. They are considered non-renewable after that time.

Having a research credit grant of this type is independent of writing a grant proposal to a funding agency such as NIH or NSF. If a cloud research credit grant is secured by you en route to writing a proposal to an agency like NSF: You may wish to reference that grant in your proposal as a means of defraying computing costs in year one of the proposed effort.

Research credit cloud accounts are distinct from paid accounts secured under the UW BAA with AWS and Azure. Those accounts have legal terms associated with them (see respective overview pages herein: Azure overview and AWS overview.

Federally funded research

FAQ

Q: How fast/powerful is a given instance?

A: You can read about instance types on the vendor’s cloud console and/or information pages. For example AWS EC2 instances are written up here.
From the Linux command line you can issue commands such as:

% lscpu
% cat /proc/cpuinfo

It is possible that (from evidence circa 2012) a certain degree of CPU type heterogeneity can introduce on the order of 10% performance variation.