RAPTOR

FIU currently has a moderate-size HPC cluster with 3,000+ CPU cores dedicated for FIU researchers to run scientific code [9]. It is a traditional parallel platform that provides hardware and software resources to domain scientists to run parallel workloads. The HPC cluster does not adequately address the diverse needs of the domain science applications, nor does it provide adequate support for big data, machine learning, and AI applications that are becoming more prevalent. At FIU, the coastal resilience and sustainable environmental research requires on-demand computing resources, facilitating machine learning, data processing, and data analytics on both CPUs and GPUs. We envision RAPTOR will in the future be incorporated as an integrated cyberinfrastructure supporting real-time computing with resource allocation enabled by Chameleon so that the platform can be connected with a rapidly deployable sampling system that can assimilate and transmit data in real-time to facilitate actionable intelligence and drive adaptive environmental monitoring decisions for aquatic ecosystems. With on-demand processing of real-time data, a networked array of autonomous platforms can be deployed in addition to RAPTOR to efficiently monitor aquatic habitats and respond as conditions change.  

RAPTOR will be integrated with Chameleon (as a part-time associate site) and Open Science Grid (contributing to the open science pool) as coordinated shared computing resources for multi-campus and national resource sharing.  

Chameleon is a deeply reconfigurable platform that allows provisioning of bare-metal compute nodes while giving the researcher full control of the software stack allowing customization to meet the research needs. Although the original goal of Chameleon is to provide a production testbed for computer science researchers to conduct experiments in systems and network research, RAPTOR will take advantage of Chameleon infrastructure-as-a-service (IaaS) capabilities and reconfigure the compute nodes (and their interconnect) on demand to meet the domain application requirement. 

OSG allows institutions across the country to share resources (by deploying idle CPUs for OSG workloads). This can significantly expand compute and data processing capabilities at FIU for applications, such as the hurricane loss models and the proteomics search. OSG also operates the StashCache data federation [10], a content distribution network designed for scalable and transparent distribution of data for opportunistic users of the cyberinfrastructure. For example, proteomics studies require computational tools and algorithms, and handling large-scale proteomics data. These tools and algorithms can run on diverse computing platforms, ranging from stand-alone servers, to specialized systems equipped with GPU accelerators or other types of co-processing units, to full-scale parallel and cloud computing platforms. They can take advantage of the OSG’s federated distributed computing and data distribution services. 

RAPTOR is poised to succeed, not only by introducing new computing modalities at FIU to support its researchers, but also benefiting a broader federated community of researchers and scholars sharing resources through OSG and Chameleon Cloud. The integration of knowledge, tools, and modes of computation proposed by RAPTOR has the potential to demonstrate novelty from integrated research and learning approaches. RAPTOR will be available as an open resource to provide new learning and experiences to students and help them to become the next generation of transdisciplinary researchers and professionals. FIU will use RAPTOR to further develop research and learning experiences for its diverse community of students.