Fujitsu BestPlace is Data-driven to the Best SAP Placement

E3: Fujitsu BestPlace promises to find the best location for each SAP instance in an environment. How does that work in practice?

Hendrik Müller, Fujitsu: In practice, we talk about "workload placement", i.e. the optimal distribution of SAP instances in hybrid multi-cloud environments. Placement decisions have become increasingly complex in recent years. In addition to traditional on-premises, hyperscalers and SAP itself offer a variety of operating environments. Gartner describes this as an "explosion of workload deployment options" and warns that incorrect placement can result in high operating costs, performance problems, or even jeopardize business continuity. Workload placement must be based on data and requirements. Because requirements and SLAs change over time, this is a recurring task. Gartner estimates that 85 percent of placement decisions made today will not be optimal in five years.

E3: What does optimal mean in this context? How can SAP systems be optimally placed in the cloud?

Jörg Niopek, Fujitsu: As is often the case in our industry, the answer depends on the specific requirements. What is ideal for one customer may be inappropriate for another. Each customer and each system has unique strategic, organizational and operational requirements, some of which may be conflicting. For production systems, optimal placement is designed to maximize availability and performance while minimizing cost. A critical factor is the network traffic between systems. In an on-prem environment, this is not a cost driver, but in a hybrid cloud environment, it can be a significant cost driver. Clever placement can minimize these costs.

E3: Is the problem truly complex enough to require an algorithm?

Müller:‎ Yes, because the number of possible placement combinations is enormous. For example, if we assume twelve SAP systems, 54 regions in Azure, two on-prem data centers and two other placement options within Rise with SAP, then there are 1.45 trillion potential solutions. While many are not valid due to constraints, it remains an optimization problem that cannot be solved manually in a reasonable amount of time. There is also a human component: If you ask three consultants for the optimal solution, you often get three different answers. Microsoft, AWS and SAP also have their own perspectives. To counter the risk of a biased solution, Fujitsu takes a data-driven approach that focuses on the customer's specific workloads and requirements.

E3: How do you solve this problem?

Müller:‎ We automate the solution finding process. We use optimization algorithms to do that. Workload placement is a separate area of research that often relies on heuristics. However, heuristics are not suitable for complex business applications because of the complexity of implementing placement conditions. Instead, we rely on genetic algorithms, which work on the principle of "survival of the fittest". The algorithm simulates a mini-evolution: it generates a number of random solutions, evaluates them, and evolves them through mutation and recombination. This leads to optimized solutions over several generations. A selection process ensures that promising solutions and some suboptimal solutions are further developed in order to comprehensively search the solution space.

E3: And who evaluates the solution afterwards?

Müller:‎ The evaluation is carried out using a fitness function. The better the "fitness", the more suitable the solution. The customer's constraints are taken into consideration by a penalty function. Instead of immediately excluding invalid solutions, they are penalized in order to avoid local optima and find the global optimum. The result is an optimal hybrid solution that we present to the customer and, if desired, compare with alternatives such as a pure on-prem or 100 percent Azure or RISE solution.

E3: What concrete results and recommendations does BestPlace provide?

Niopek:‎ BestPlace provides a reference architecture for an optimal hybrid SAP infrastructure. This includes cost-optimized placement, appropriate sizing and a detailed cost comparison for compute, storage, network, licensing, and operating costs. All the data collected is available to the customer for later analysis.

E3: You mention a data-driven service. Where does the data come from?

Niopek:‎ In order to optimize our service, we need to understand two key aspects: the workload of each SAP instance and the requirements of the associated systems and organizations. Immediately after the project kick-off, we start with a comprehensive measurement. We use software in the customer's environment that is connected to all relevant systems via standard SAP interfaces. This allows us to collect key data on capacity, resource consumption, and usage patterns. At the same time, we conduct two workshops with the customer: In a strategy workshop, we match SAP systems to appropriate requirement profiles and define strategic, operational, and organizational needs. In the cost workshop, we refine our cost model together with the customer to take individual factors into account. This enables us not only to analyze infrastructure costs precisely, but also to compare operating and location costs in a targeted manner.

E3: What kinds of customer challenges are you seeing?

Müller:‎ At the end of the day, it's about meeting SLAs for the customer and the business. Capacity management means meeting requirements cost-effectively without compromising performance. To do this, we need to understand the technical, organizational and business requirements. From a technical perspective, this includes resource buffers for sizing, system availability, data throughput requirements, and backup and disaster recovery strategies. At the same time, we must ensure that the potential operating environments can support the specific requirements of the systems. On the organizational and strategic side, the importance of IT to the business and the level of innovation play a role. It is also important to understand what expertise is already available in the organization, for example, in cloud architecture or operations, as this will affect subsequent operations and efficiency. From a business perspective, we need to answer the following questions for each potential target environment

E3: How does BestPlace differ from other tools?

Niopek:‎ BestPlace is not a tool, it is a consulting service. There are many monitoring tools, but our added value lies in the combination of measurement and analysis. We interpret the data and derive specific recommendations for action. Where it makes sense, AI supports our analysis, for example by providing publicly available information first.

E3: How does RISE achieve optimal placement of SAP workloads?

Müller:‎ Even though RISE with SAP goes beyond a pure IaaS offering of hyperscalers, our principle remains the same: The focus is on the requirements of the customer's workloads. On this basis, we check in which scenarios a RISE contract makes sense. A key cost factor is the FUE (Full-Use Equivalent) requirement, which we record in detail. We then analyze which additional infrastructure components are required, such as higher availability classes, extended disaster recovery requirements, or longer backup retention times. As with all placement options, we begin our evaluation with publicly available information. If the customer already has individual quotes, these can be taken into account. In this way, we ensure that the workloads are placed in the best possible way, both technically and economically.

E3: What about licensing?

NIopek: The more diverse the requirements, the more complex licensing becomes. The number of systems is less important than their heterogeneity. The cost of a BestPlace project therefore depends on the number of defined requirement profiles. This approach is particularly scalable for medium and large SAP environments, as additional systems can be assigned to existing profiles at no additional cost. Our goal is to minimize subsequent operating costs. Increased planning effort in the design phase usually leads to cost savings and prevents unexpected additional expenses—the so-called "cloud bill shock". Experience shows that correction costs are significantly higher than design costs, which is why detailed planning is worthwhile.

E3: What other use cases are there for AI-based SAP consulting?

Müller:‎ Our data-driven services can already predict workloads and identify performance anomalies. Since we have been actively researching and publishing in this area for more than a decade with our own lab, we are always open to new use cases. Where there is real added value, we integrate the models we develop into our consulting services, such as the SystemInspection Service for SAP Solutions or BestPlace. Our latest Large Language Model (LLM) explains SAP transactions and facilitates performance analysis. In the future, LLMs in the form of digital administrators could prepare data-based decisions and alleviate staff shortages in IT operations teams.