The steep learning curve and intricate toolchains associated with FPGA design on Amazon’s cloud can impede development. CETIC’s proposed architecture provides a simplified pathway for integrating and deploying FPGA computations.
Field-Programmable Gate Arrays (FPGAs) are increasingly utilized for hardware acceleration in applications requiring high bandwidth and computational throughput. Use cases span various domains, including aeronautical simulation, climate modeling, artificial intelligence, Big Data analytics, and genomic research.
FPGAs are configurable integrated circuits that enable parallel execution of processing tasks. This differs from traditional Central Processing Units (CPUs), which execute instructions sequentially. Parallel execution in FPGAs can accelerate complex computations.
Amazon Web Services (AWS) provides FPGA instances that integrate CPUs and FPGA boards. These instances facilitate FPGA-based processing, reducing the need for significant hardware investments and leveraging cloud scalability. This solution is applicable for organizations seeking hardware acceleration and cost optimization.
Within the BAMetagenomes project, CETIC developed an FPGA architecture designed for efficient utilization of Amazon FPGA instances. The architecture is structured around a "Load-Compute-Store" model :
Amazon’s FPGA cloud offers a flexible and scalable environment for FPGA-based processing. This infrastructure reduces hardware constraints, enabling experimentation, testing, and optimization of FPGA designs across multiple application domains. CETIC currently demonstrates two services on this architecture : integer multiplication and DNA sequence hash computation.
Our team is available to provide demonstrations and discuss potential collaborations in FPGA-based solutions. For inquiries regarding this architecture or related topics, please feel free to reach out to us by contact the authors or sending email to : info@cetic.be