ÓÐÁϺÐ×ÓAPP

NSF’S grant systems modernization would require the migration to the cloud of a mission-critical suite of more than 20 tightly coupled monolithic legacy systems—a foundational step for achieving parity with current industry standards in security, reliability, failsafe redundancy, and readiness to incorporate new and emerging technologies down the line. Other central priorities of the modernization included conducting a full, user-centered redesign of NSF’s proposal submission portal, and enhancing the program’s ability to scale user capacity up and down as needed to accommodate surge periods. All of this work needed to be undertaken during normal operations, and thus would have to be planned and carried out in such a way as not to disrupt scientists’ ability to securely access their NSF accounts from anywhere in the world, especially during surge periods which predictably take place in the lead up to major program application submission deadlines.

Today, more than 90% of NSF’s grant management operations have been relocated to the cloud. Everything is deployed with at least two backups to ensure—along with automated scaling capabilities and frequent health checks—that core services are buttressed by the redundancy necessary to virtually eliminate outages.

The automated testing, deployments, and enterprise monitoring employed by mature DevSecOps pipelines combine to multiply efficiencies and speed delivery. For example, the automated build processes that we implemented reduced production deployment times from more than 30 minutes to less than three, saving the equivalent of 28 development days each year. Pipeline standardization empowers application teams to configure for scaling events without waiting for additional approvals, saving time, and reducing stress around deadline surges.

Our pre-surge stress tests provide further certainty around deadline periods. For example, when performance tests showed utilization reaching 85% prior to the 2022 application deadline for the , we preemptively increased CPU and memory utilization to guarantee resiliency during the surge.

NSF’s new proposal submission application, created using agile development processes and user-centered design principals, has seen an 87% improvement in user satisfaction scores compared to its predecessor. To reduce dependencies and allow for development and deployment flexibility, we built the new app atop 11 microservices residing in independent domains. By migrating components of the application from one database to another, we achieved more than $100,000 in annual cost savings while improving per-transaction performance by up to 50%.

With an eye toward the future, ÓÐÁϺÐ×ÓAPP continues to help NSF monitor tech trends and industry standards in order to plan for the adoption of new and necessary technologies as they emerge and mature. To date, we have successfully modernized and deployed more than 80 applications and services to the cloud and managed the strategy to decommission the associated legacy applications. As the federal government’s largest provider of AI services, we are leveraging that expertise to assist NSF with AI system development and ensure that the AI tools and processes it adopts are technically sound and directly in line with its mission and business goals.