March 1, 2021 — Credence, under the Defense Logistics Agency’s (DLA’s) Procurement Integrated Enterprise Environment (PIEE) hosting support contract, successfully migrated a mission critical eCommerce ordering application, FedMall, one week ahead of schedule in March 2020. As a Managed Service Provider, since then Credence has been managing the FedMall application for the DoD, Federal, State, and authorized local agencies to search for and acquire products from government reserves and commercial sources. Buyers have access to more than 29 million individual items of supply, including centrally managed DoD and General Service Administration (GSA) assets, as well as Commercial Off-The-Shelf (COTS) products. FedMall has approximately 110,000 registered users, 732 commercial suppliers, 53.3 million commercial items, and processes millions of orders annually. Credence, using the provision of Infrastructure-as-a-Service (IaaS) cloud computing resource, collaborating with Amazon Web Services (AWS), migrated FedMall to the AWS GovCloud environment.
Problem Solution/Definition:
DLA hosted FedMall in IBM managed data centers and had issues monitoring resource utilization, virtual machine provisioning, and meeting the needs of Agile software development teams including issues with application uptime and costs. Our client requested insight on the future trajectory of the Fedmall application.
Proposed Solution & Architecture:
Credence migrated FedMall to AWS GovCloud infrastructure using “lift and shift” philosophy of migration. We utilized AWS CloudEndure service to migrate the virtual machines from IBM data centers to AWS GovCloud. KPI’s were calculated from S3 files, through code implemented in glue. Multiple regressions and logistic regressions were fit to the calculated KPIs compiled over years of data to create predictive models. These models were used to predict the health and use of the FedMall application months in advance.
Outcomes of Projects & Success Metrics:
The application migration was seamless to the end users with no unplanned downtime of the application. Furthermore, post migration Credence performed Well Architected Review of the application and has implemented various recommendations in rearchitecting FedMall and utilize many cloud-native services. On application side, projections of user base changes and purchases on the application were possible 3 months in advance with 95% confidence.
Describe TCO Analysis Performed:
AWS pricing calculator was used to determine cost of infrastructure used to host and run code and storage. We have successfully brought down the TCO of FedMall by adopting cloud technologies.
Lessons Learned:
Through building this application, we learned how to efficiently define advanced AI ensemble models in scalable server less AWS infrastructures, like lambda container and glue. We also learned how predictive models can be used as anomaly detectors.