Enterprises are increasingly challenged to adapt to the rapid data growth and real-time processing requirements in data centers and the cloud. A more and more common solution is to utilize hardware accelerators that can ease big data analytics by bringing compute tasks from hours to minutes/seconds, or reduce the latency of real-time critical applications down to milliseconds. But then a new question arises: how can hardware accelerators be included without disrupting the workflow? Xelera Suite is a hardware-agnostic, on-demand scalable and easy-to-deploy acceleration platform which provides push-button access to the most advanced GPUs, FPGAs and AI ASICs in data centers and public clouds.
Xelera Suites Compute Hardware Interface provides a single API to FPGAs, CUDA GPUs, and OpenCL devices. The North-bound API remains the same regardless of the device type. The Compute Hardware Interface is an efficient shim which does not add any performance loss.
Computing tasks are distributed automatically and transparently across accelerator cards, CPU cores and server nodes. Xelera Suite provides a high scale-out efficiency using a carefully engineered overlay network service.
As a unique differentiator, Xelera Suite guarantees an ultra-low latency at system level: from the data source through software infrastructure and virtualization frameworks to the accelerator. This is achieved with a technique called system-wide slicing.
Xelera Suite provides a set of rich libraries with FPGA and GPU device software to enable hardware acceleration in a large variety of software applications.
Xelera Suite can connect different type of hardware accelerators such as FPGAs and GPUs via a high-throughput and low-latency peer-to-peer bridge. This allows users to build highly efficient hybrid acceleration platforms.
Xelera Suite manages large repositories of acceleration firmware automatically. It detects the available accelerator hardware and selects the adequate firmware binaries according to the hardware type and application configuration. Xelera Suite can be deployed via Docker, virtual machines or bare-metal.
Xelera Suite provides monitoring, logging and reporting through a comprehensive dashboard. Alternatively, the workloads can be controlled via an API for 20+ programming languages.