Cloud Digital Repo Optimization


The goal of the project is to scale down the CloudFormation templates for deploying the Hyku digital repository application. We have attempted to reduce the cost of running the Hyku application with a base level of performance, essentially reducing it to the minimum viable scale. We have accomplished this by changing these templates and their configuration parameters to use less instances at smaller sizes. After evaluating a number of different options for reducing the base cost, including using other AWS offerings, we have settled on a number of parameters that work well at the base level of performance. In testing these changes, we used a qualitative method of testing the functionality of the existing feature set on the original deployment and comparing that to the functionality of the new deployment. We have seen no changes in functionality from the original deployment.

The cost reduction we see with these reduced instance sizes is to about one third of the original cost, resulting in massive savings given that the original cost of running the application was about $800-900 a month. The new cost of running our modified templates with the parameters we have tested is about $300 a month. Given that the original feature set is still functioning as it was before, we believe that we have achieved a satisfactory reduction of cost from the original deployment, and therefore have accomplished the goal we set out to complete.

We provide documentation on our process and the changes we made, including on how to reproduce in the future the changes we have made. Since the templates require some level of maintenance, this documentation is vital for deploying them in the future. The documentation provided by the report gives future maintainers the ability to quickly get up and running with the potential problems encountered when working with the templates, and gives future groups the insight to predict the kinds of challenges they will face when working on the Hyku CloudFormation templates.



CloudFormation, Hyku, Hydra-in-a-Box, Samvera Labs, AWS, Amazon Web Services