Workload Management and Performance

In this chapter, we specify the business requirements and propose the solution concept for workload management and performance. Depending on the underlying algorithms, training of artificial intelligence models can have high requirements for hardware. Inference calls must have very short response times. Thus, measures must be put in place to resolve those challenges. Especially, consumption of computing resources for artificial intelligence must not slow down the transactional processes in ERP systems as they are critical of running the enterprise. Consequently, in additional to the already proposed solution architecture, which scales from embedded to side-by-side artificial intelligence approach, we suggest further performance optimization techniques.