CADRE Project Director Jamie V. Wittenberg presented a virtual briefing entitled “Empowering Data-Driven Research through an Open, Accessible Data Infrastructure” at the virtual CNI conference last week.
You can watch the presentation here:
We encountered a great question from CNI Executive Director Clifford Lynch during the presentation, and CADRE’s technical lead, Valentin Pentchev, wanted to take some extra time to respond in-detail.
Q: How does CADRE handle the provisioning of analysis from a computational standpoint? If users come onto the platform with things they want to explore, it would take quite a lot of computing cycles. How does CADRE manage that?
A: CADRE is working to bring overall computational costs down in a few ways.
First, CADRE presents a single list of datasets to all participants. Costs for those datasets are significantly reduced when compared to individual implementations, and that gives us the ability to use more computational resources at a much lower price.
Additionally, by building a cloud platform, we can default to serverless technologies wherever possible. Systems that aren’t utilized can remain dormant and provide essentially no cost, while also having the ability to scale well beyond initial capacity when the need for specific services arises. The platform can also allow users to continue their work outside of CADRE with their own on-premises resources or in the cloud, since CADRE uses standardization of workloads, loosely coupled infrastructure systems, and containerization of all major processes.
There is the added potential of CADRE’s planned tiered-pricing model, where users know their usage limits. For example, the free tier we plan to offer could provide limited resources and storage based on the amount of time and space used, while offering all other CADRE benefits. A second tier may have more resources and storage available, and a third tier would offer even higher limits. We plan to incorporate a capability for users to bring their own resources to the cloud to continue to run our workflows at their own expense or on their own hardware on-premises.
The factors stated above will find the most success if CADRE’s collaborative nature persists. As Jamie said in her presentation, the more partners willing to contribute resources on the platform, the less-expensive the platform will become for everyone while offering more.