Digitizing processes requires a robust data architecture and infrastructure for data storage. “A robust data architecture is essential,” explains Glenn Fitzgerald, CDO Product Business at Fujitsu EMEA, at the start of our exclusive E-3 interview. “Especially when you need both SAP Hana data and non-SAP data to be available on a universal platform for all business needs. This is where the system inspection service comes in really handy.” For users, there are two crucial aspects: data privacy, and data security within the context of overall IT security. This solution uses artificial intelligence methods to facilitate data analysis in order to detect anomalies. “In one specific instance of collaboration with our partners, we defined and installed a suitable IT platform on the basis of Kubernetes, Ceph Cluster, Fujitsu Primergy servers and NetApp NFS servers,” says Glenn Fitzgerald, describing the symbioses in the SAP community.
Whether a company succeeds in the digital transformation depends above all on how well or how poorly it manages its data. Progress is only possible with a consistent data and analytics strategy. That’s why many companies are currently making structural changes to become a “data-driven company” – i.e. a company that consistently uses its own data resources to open up new opportunities and possibilities for their business processes.
“During a digital transformation, data must be prepared, enriched with information, consolidated, and correlated,” says Thomas Herrmann, Manager Business Development SAP at NetApp for EMEA. He organized a conference for SAP customers this September, which was attended by SAP itself, Cisco, Fujitsu, Red Hat and Amazon/AWS. The conference demonstrated how IT providers for data management in SAP systems can interact in a complementary way. Cloud computing was a key topic, with special attention paid to the hybrid cloud and the challenge of S/4 conversion. Thomas Herrmann put it this way at the NetApp event: “So, data management plays a critical and important role in all corporate digitization initiatives. Data management is the sum of all measures necessary to collect, store and provide data. Once digitization is complete, all key business processes will be based on data, which must be optimally managed in order to achieve the optimal effect.” His colleague Robert Madl from Cisco adds: “Data management is definitely an important criterion for success. After all, digital transformation with SAP involves the digitization, optimization and automation of business processes.”
“Hybrid data architectures are feasible. The challenge lies in the interdependencies between the systems.”
Robert Madl, Global Strategic Partner Executive, Cisco
Data management is currently in demand everywhere, as Cisco manager Robert Madl explains: “For example, if I automate processes in production or warehouse logistics using sensors – keyword IoT – a lot of new data is generated that has to be handled in a completely different way than traditional ERP data. If machines are controlled on the basis of this sensor data, for instance, reliable transmission with correspondingly low latency is critical. If you collect sensor data for a “big data” analysis in connection with data from the ERP system, it’s no problem to store this data in a hybrid manner, i.e. distributed in a data lake in the cloud, while keeping the ERP databases on premises or in a colocation data center. But making the right data management decisions is critical to the success of digital transformation projects.”
By 2024, 93 percent of companies in Germany will use their data to drive revenue growth. 42 percent will even use data as a significant source of revenue. This is according to the new study “The Multi-Cloud Maturity Index,” which was conducted with around 3000 business and IT decision-makers in the EMEA region. “The past decade has shown that almost everything around us is data driven. What’s more, data has become a key corporate asset and, if used correctly, can be a significant contributor to a company’s success,” says Glenn Fitzgerald of Fujitsu in a discussion with E-3 editor-in-chief Peter Färbinger. Fitzgerald adds: “We are currently living in a world of unstructured data, data silos, exorbitant data growth, and increasing data complexity. This makes the management of this data all the more important to the company’s success. How sound is a company’s data management strategy to ensure that it can respond quickly to market demands at all times?”
“Intelligent enterprise that achieves a holistic view of S/4 data and non-SAP data.“
Glenn Fitzgerald, CDO, Fujitsu PBL Europe
There is a key question that must be answered: What methods and tools are needed to successfully use data of any kind? Technologies such as artificial intelligence and machine learning offer possible solutions, says Glenn Fitzgerald, who further explains: “On the one hand, they can significantly improve the quality of the data, and on the other hand, they enable you to detect errors during data collection. This can be supported by automated machine learning acquisition. Supplying business processes with optimal and qualified data –and at the right time–is key to a company’s success. One of the main goals is to meet the customer on their terms, identify their challenges and then deliver a solution, which we can accomplish using a number of different techniques, methods and tools. Our goal is to support the customer and work with them to develop their intelligent enterprise.”
What are the criteria for data storage in an intelligent enterprise? If the ERP system is on premises, the data should also be on premises? If the ERP system is in the cloud, the data should be there too? Is that right? Thomas Herrmann: “That’s not so easy to answer because there are several factors at play: network speed, i.e. sufficient bandwidth to the cloud, location, and distance to the nearest backbone. Also: real time access or batch processing? That is, what are my SLAs in terms of response times, etc.? When it comes to realtime processing of data, the data should of course be located in the same place it is processed. With Hana that would then be in-memory computing. Whether that is in the cloud or on premises is secondary.” And Robert Madl from Cisco clarifies: “Hybrid data architectures can be implemented, of course. The challenge is to understand interdependencies between systems. Often SAP landscapes have developed organically over decades, and custom code is implemented everywhere – the creators of which may no longer even be in the company. This often leads to dependencies between systems. For example, maybe one system can access the database of another system directly – or it can make a call to the other system, which in turn grants access to the data tier. In this case it’s important to understand which systems depend on each other and how. In other words, what bandwidths are needed and how time-critical is this communication, i.e., what are the maximum latencies allowed to provide the necessary data in a timely and complete manner?”
“The flood of data that comes with digitization requires an archiving concept.“
Thomas Herrmann, Manager Business Development SAP, NetApp
Processes and algorithms
Ultimately, it is a matter of ensuring that the business process mapped in the SAP systems functions efficiently, regardless of where they are running. “You just have to be aware that when you’re distributing SAP systems across multiple sites, you’re going to have higher latencies and lower bandwidth between sites, and you have to take that into account during your premigration planning. This is where AppDynamics can be really useful, as it automatically analyzes and visualizes these dependencies between systems and makes them available for planning,” explains Robert Madl.
What does the Cisco manager think about the best way to store data? “That depends on the type of data and how it’s used. With databases like SAP Hana, it makes sense to have the data close to the computing resources,” explains Robert Madl. “While Hana is an in-memory database – meaning that data is kept in the server‘s memory – that only helps with read transactions. Write transactions are only confirmed when the data has been written to the persistence layer, otherwise known as the data storage system. In this case it is critical to have fast IO between the server and storage location for optimal application performance.”
The biggest performance boost for OLTP applications came with the introduction of flash memory. In analytics scenarios (typically OLAP), the performance impact at runtime would be lower with in-memory technology, since the data is already available. But it would take a very long time to boot these systems if the data cannot be loaded from a local data store into RAM. Decentralized data storage can be really useful for ‘big data’ analyses,” says Robert Madl, who goes on to explain: “For example, if you have multiple data lakes built on Hadoop close to the data source or sensors, you can – for example, with the MapReduce algorithm – pre-aggregate data for analysis in a decentralized and iterative manner and then transfer only the necessary information to a central system for further processing.”
What are the advantages and disadvantages of hybrid data management? Here again Robert Madl: “There are three factors that must be optimized: time, cost and complexity. The place where the data originates is not necessarily the place where the data is used. Transmitting data over long distances costs money and takes time. But having many different storage locations increases complexity. In digital transformation projects, it often proves useful to define a minimum time requirement and an upper cost limit, and then to optimize the complexity dimension first. For example, in a smart factory project, time often dictates how much edge computing is needed, the cost of transmission dictates the degree to which sensor data must be pre-aggregated, and complexity is ultimately the deciding factor for feasibility and overall success.”
S/4 and data conversion
In S/4 conversion projects, data management and data storage are key factors that determine the cost and overall success of the project. How can you guarantee high success rates and low data costs for existing SAP customers? “Together with our customer, we are building a truly intelligent enterprise by taking a holistic view of SAP S/4 Hana data and non-SAP data,” says Fujitsu manager Glenn Fitzgerald in describing the challenge. Precisely how the data is managed and stored depends on the company’s own business processes. “This is where Fujitsu supports its customers with its co-creation approach. In essence, this approach involves conducting a workshop based on the Fujitsu Human Centric Experience Design specifications. We work closely with customers, technology partners and our experts to develop an optimal approach, along with a proof of concept and a long- term plan to overcome specific challenges and continuously optimize IT,” says Glenn Fitzgerald, speaking from experience of many successful projects.
Digitization and the flood of data
Existing SAP customers will likely see their volume of data continue to grow, and with it the cost of data management. “The flood of data that comes with digitization requires an archiving concept,” explains NetApp manager Thomas Herrmann at the end of the E-3 interview. “The first step is to determine what data must be archived due to legal requirements, what data you want to archive, and what data needs to be retained for a specific period of time. Modern data archiving is based on the cloud. All major cloud providers offer an archive tier for object storage. These tiers are increasingly becoming the preferred destination for backup data with long-term retention requirements. This includes all major archiving packages from AWS, Azure and GCP. Cloud archive solutions are the most cost-effective object storage tiers available today, and they can be scaled to petabytes of storage as the volume of archived data increases. NetApp Cloud Backup, for example, provides a comprehensive service for long-term protection of your data in heterogeneous environments, whether in the cloud, on premises, or in a hybrid combination of these platforms. NetApp Cloud Backup supports the archive tiers from the cloud providers above as destinations for your long-term backup and archival data.”
Data and workloads
Cisco manager Robert Madl has another tip for SAP customers: “SAP workloads don’t usually exist in a vacuum. This means that an infrastructure that is optimal for SAP Hana should be optimal not only for Hana itself, but also for all other workloads, so that you don’t have to build an additional IT management silo for this one workload. There are around 200 reference architectures for the Cisco FlexPod that show how to reliably run workloads on it – not only SAP workloads such as Hana, but also web services, which are often one of the mapped business processes supported by the SAP system. FlexPod XCS is the new version of FlexPod optimized for multiple clouds, which expands on these reference architectures to include scenarios where you can outsource and connect services to the cloud without additional management effort.”
This is the first part of a five-part series! If you would like to read the second one, click here.