Beyond many-core: Commercial workload acceleration in high-end systems

Summary form only given. Keeping up with Moore´s law is becoming more and more difficult. The industry is acknowledging this and increasingly turning to acceleration to solve this challenge. In my talk I will present different acceleration solutions (GPU and FPGA based) that are used for commercial workloads and their advantages and drawbacks based on the workload characteristics. I will discuss additional constrains that play a key role in this environment like security and availability. The Internet of Things revolution with massive rich sensor data to analyse and process is providing another area of great interest and challenge at the same time for the industry. I will show a novel approach under development based on non-von Neumann architecture to address this issue and give my outlook on commercial acceleration needs and directions in the future.