This series is available as PDF download (English, Spanish and French) at the bottom of the page.
Professor Hasso Plattner chose his first partner for in-memory computing database Hana poorly. Together with Intel, he and his team at the Hasso Plattner Institute (HPI) in Potsdam, Germany, came up with a concept for in-memory-based databases. Eight years later, it became clear that the concept was right – but the hardware and the concept of appliances were a poor fit.
After the very first installations of Hana on Intel, experts were already convinced that performance, scaling, and virtualization would not reach their full potential with this combination. However, Intel would not or could not improve its hardware. Maybe Intel also believed that their connection to SAP made them unimpeachable. And for a while, it really did seem like it. Hasso Plattner and former SAP CTO Vishal Sikka long fought against transferring the Hana database to the IBM Power architecture. Even though the community already knew that HoP, Hana on Power, was the far better solution, SAP long denied IBM its Hana certification.
Even a quite exotic HP subsidiary like SGI became certified for Hana almost overnight, while IBM and its Power architecture had to watch from the outside. Meanwhile, Hana problems got bigger and bigger on Intel’s platform with every passing day. Appliance sizes were not what SAP customers imagined, and the virtualization with VMware was long not suitable for daily operations.
It was common knowledge in the SAP community that Hana operating system Suse Linux was also available on the IBM Power architecture. Switching would therefore have been no problem. Finally, SAP caved, and IBM was able to officially prove that its Power server architecture was significantly better suited for Hana than Intel processors. Of course, IBM also got rid of the one-size-fits-all appliance model and gave customers the Hana servers they had imagined all along.
SAP’s and HPI’s attitude towards IBM Power changed as well. The Hasso Plattner Institute started offering an online course on the future of computing on May 1st, 2019. “Future of Computing -IBM Power 9 and beyond” spanned over four weeks and was available free of charge on the IT learning platform openHPI.
IBM Power at HPI
This online course was organized by HPI Professor Andreas Polze, Operating Systems and Middleware, Hildegard Gerhardy from the IBM Academic Initiative Europe, and Wolfgang Maier, director of IBM hardware development in Boeblingen, Germany. “We show participants different approaches to tackle the challenges of digitalization, especially the exponential growth of data,” Polze explained.
He highlighted the fact that the information storage capacity per capita has nearly doubled every 40 months since the 1980s. “Not to mention that many experts actually expect the global data volume to reach 160 zettabytes in the next five years,” added Gerhardy.
“More and more sets of unstructured data are generated, e.g. through the Internet of Things, and have to be analyzed. It is therefore necessary to consider different approaches in software development,” said Wolfgang Maier. This includes the availability of microservices, container solutions, and cloud-based applications.
Furthermore, IT departments need new basis technologies, like hardware accelerators, artificial intelligence, and blockchain, to cope with huge data volumes. Polze and his colleagues contrasted this trend of new data analytics in so-called “systems of engagement” with technologies of traditional “systems of record”.
“Reliability, high availability and serviceability of systems require highly developed and sophisticated hardware, operating systems and application-independent programs to make successful transactions on huge scales possible,” explains Polze. The HPI online course focuses on the future of computing with IBM Power systems. IBM cooperates with more than 200 member companies in its Open Power Initiative, including Google, Samsung and Nvidia, to create diverse innovations in software and hardware.
Big data and DB architecture
In his Sapphire 2019 keynote, Hasso Plattner also highlighted the enormous data growth and the necessity to leverage the right server architecture and databases to quickly find solutions. In his mind, the migration to Hana should have taken SAP customers about three years. However, the database release change took longer than expected; which can partly be traced back to SAP’s early commitment to the Intel platform unsuitable for Hana.
Hana development at the Hasso Plattner Institute at the university of Potsdam, Germany, focused on Intel and the x86 architecture of its Xeon processors. However, the IBM Power architecture turned out to be the better foundation for Hana. SAP gave up resisting in 2014 and allowed HoP – Hana on Power.
In-memory computing database Hana has many exceptional advantages compared to traditional SQL databases; not because SAP is better than other companies, but because Hasso Plattner dared to start from scratch. Without relics and legacies, the teams at HPI and SAP’s headquarters in Walldorf, Germany, were able to begin anew.
In Potsdam, students of Plattner researched and programmed together with Alexander Zeier. Mathematician and former SAP CTO Vishal Sikka gave Hana the last finishing touch.
Plattner, Zeier and Sikka not only knew Intel’s processor architecture like the back of their hand. All three of them also believed that only this powerful processor would be suitable for their Hana database. In 2014, they officially realized that this was a misconception.
IBM Power for big data
“Power 8 and also Power 9 were specifically developed to process big data volumes. Power has four times more memory and five times more cache than Intel x86. It also has higher performance than x86 – twice as much per core in benchmark, and four times as much with real customer workloads. Furthermore, it has higher flexibility with PowerVM and high reliability through redundancy, especially with Power Enterprise servers,” explains Andreas Klaus Span, Director and Business Unit Executive for SAP Hana on Power.
In general, Andreas Span believes that because Power comes from the enterprise realm and therefore has a different architecture, it is better suited for Hana than Intel x86 – which comes from the commercial sector and never had to be more than just good enough.
Power as differentiating factor
Since 2015, Hana has been available on IBM Power systems with IBM’s innovative Power 8 architecture and processors (currently Power 9). IDC’s market analysts are convinced that Power systems are a powerful differentiating factor for Hana and S/4. Power was specifically developed for data-intensive workloads like Hana and includes integrated, SAP-certified virtualization as well as numerous features to improve reliability.
Andreas Span knows that because of weak virtualization, the increasing number of restrictions and security mistakes, and low capacity of x86 memory DIMMs, TCO approaches are not as economical as some might think. He said, “Furthermore, the Hana database and the data volume is growing exponentially. Considering these factors, Power is not only the more high-quality, stable platform, but also the more cost-efficient.”
The flexibility of IBM Power systems allows for the simultaneous operation of various environments. This means that customers can for example use unused capacity from their productive environments for development and user acceptance tests. Compared to other architectures, IBM Power achieves higher efficiency from more distributed resources.
IBM Power furthermore gives customers the reliability they need for critical Hana workloads. Because of reliability, availability and maintenance features, the Power architecture is uniquely suited for Hana implementations. This combination moreover supports a variety of different mechanisms, tools, and processes, including high-quality support for redundancy and replication.
Tailored Datacenter Integration
As always, everything depends on size. Andreas Span explains, “There may be companies, who have smaller databases and not as many applications, who could benefit from Intel. However, this is mostly only possible following a TDI approach (Tailored Datacenter Integration) – which even SAP recommends now.”
The entire Power platform is certified for Hana. “Once and for all,” said Andreas Span. “This means that we don’t have to certify every server individually, or have to have every little modification approved, like it would have been the case with appliances. Before we release new versions, like Power 9, we start testing together with SAP as early as the development phase. If everything works, we are good to go.”
SAP wants every customer on the Hana database by 2025 – and if it’s not the database, then it will certainly be the Hana platform. “For this to happen, they need a partner who can offer them a similar vision for the future,” said IBM manager Span. “We can act as such a partner. Apart from that, the code for Hana on Power and Intel has always been nearly identical – over 97 percent were the same, in fact. By now, there’s only one development department for both platforms, the code is identical, and the release times are the same.”
IDC’s market analysts are therefore right in saying, “Switching to a SAP Hana in-memory platform has gotten easier in the past couple of years. Many companies already made the first step towards Hana with a migration to SAP Business Warehouse. BW is a good starting point for a SAP Hana in-memory database.”
IDC also has a similar point of view on the market situation, “IBM is positioning itself as a Hana and S/4 expert who can offer customers a complete package – from deciding on a strategy and functional specifications to IBM Global Business Services and implementing and providing Power-based hardware on-prem and as hybrid cloud. As early as April 2016, IBM and SAP announced a digital transformation partnership focusing on innovative solutions to create cognitive expansions, user experience, and industry-specific functions with Hana and S/4. There are many reasons why IBM Power systems are an excellent platform for Hana, but the main ones are exceptional flexibility, resiliency, and performance.”
It should also be mentioned that economic benefits do not stop with TOC approaches. An early Hana positioning and implementation can companies help achieve competitive advantages.
Furthermore, it can help companies to position themselves better in the future. In this context, IBM is not only an infrastructure provider, but also consultant and companion on a journey which has only just begun. Because Hana is not a mere database – it is a constantly improving ERP/CRM platform for S/4, BW/4, and C/4.
This is the first part of a series! If you would like to read the next one, click here.