We cannot afford to overlook brainwaves

Dynatrace and JKU make Austria a digitization hotspot

Originally written in German for Johannes Kepler University in Linz.

Fully automated, self-healing and self-protecting — this is how software will be operated in the future. With this vision, Dynatrace and Johannes Kepler University (JKU) are founding a joint co-innovation research lab at the Linz Institute of Technology (LIT). The cooperation between business and science will advance application-oriented basic research in the field of software intelligence. Austria will thus become a hotspot for digitization research. Looking back at Dynatrace’s rapid development — founded in Linz in 2005, IPO in New York in 2019, around 2,800 employees, $545 million in annual revenue — it is all too easy to overlook the basis for this success: Meticulous development work to expand the company’s position as a global market leader. Dynatrace holds over 207 patents. To maintain this momentum in the coming years, the company founded Dynatrace Research a year ago, a separate research unit from its operating business. Now its head, Alois Reitbauer, is forcing the exchange with academic research. The Co-Innovation Lab at LIT sees itself as a bridge between research and business and acts as a radar system for flashes of inspiration. The cooperation is deepened by university teaching activities of Dynatrace experts and two post-doc positions.

“We have a team of around 1,000 developers who ensure that our product is the best on the market and continuously extend this lead,” explains Alois Reitbauer. Because innovation cycles in IT have shortened rapidly, new paths have to be taken in order to secure a technological lead, not just in the short term. That’s why Dynatrace now seeks close personal contact with the academic elite, who pursue research projects entirely without economic constraints and deadline pressure. “We can’t afford to overlook flashes of inspiration that might trigger a surge in innovation or even disruption,” Reitbauer says. He, therefore, looked for a digitization expert with campus experience for the job of lab director at JKU — and found him in Andreas Hametner. The Upper Austrian native completed his degree in computer science at JKU and is now returning to his former place of training.

Meeting quality is a central task of the Co-Innovation Lab for JKU Vice-Rector Christopher Lindinger. “The Co-Innovation Lab is an excellent role model for research cooperation between universities and companies like Dynatrace. As an interface between science and business, the LIT Open Innovation Center at JKU embodies an ideal place for this encounter, which provides essential impulses for the further development of the business location. JKU’s research excellence in the field of IT security, Industry 4.0, Artificial Intelligence or Big Data represents a great asset for application-oriented product development,” says JKU Vice-Rector Christopher Lindinger. The Co-Innovation Lab at LIT is the perfect place for this, emphasizes Veronika Leibetseder, Director R&D Labs Operations at Dynatrace. “Our lab on campus is designed as an open space that signals openness and transparency.” This, she says, allows students to find a place to work there at any time and experts with whom they can exchange ideas. “We want to show who we are and what culture shapes us. That’s why the lab acts as a sort of Dynatrace embassy on campus.”

Return to the roots

For years, Dynatrace has been active in the international Internet standardization body W3C and is the only Austrian company involved in the Cloud Native Computing Foundation (CNCF). One of the CNCF’s credos is to make the results of its work openly and freely available to the market. The cooperation with JKU is now a return to the roots for Dynatrace. After all, Dynatrace was founded in 2005 by three JKU graduates. The technological heart of the world market leader in software intelligence beats in the newly built engineering headquarters in Linz, which will be occupied in 2019. “For us, basic research at JKU is a kind of radar system for upcoming developments,” Reitbauer emphasizes. The 7-member Dynatrace Research department, which he heads, will move to the JKU campus and is expected to more than double in size within a year. In the research focus areas of Distributed Data Systems, Realtime Analytics, Data Science, and Cloud-Native Security, there is now a need for increased cooperation with academic researchers. “This is necessary because we know that in just a few years it will no longer be possible to manage the exponentially growing volumes of data with current instruments and methods,” says Reitbauer, outlining the scenario that research and industry foresee.

Making unimaginable data volumes manageable

It is a data explosion that Reitbauer is counting on. While an IT architecture with 100 servers used to be considered large, an environment with hundreds of thousands of servers is now quite common. Reitbauer also anticipates that data volumes could increase by a factor of 100,000 or even a million within a few years. “We should be prepared for the fact that we will soon no longer be specifying storage capacities in terabytes, but in petabytes or even exabytes.” This rapid development can be observed every day on one’s own smartphone, he said. Social media services and online retailers display their web pages in millions of individualized versions. “When I log in to my account, I’m shown my own personal browsing history, my favorites, articles viewed and purchased, links and products last clicked on, or even customized recommendations and preference from friends and acquaintances,” says Reitbauer, citing one reason for the unchecked growth in data.

Escalating complexity as a key research topic

“The digital convenience we value so much when shopping, paying, traveling, parking or banking is a one-way street whose frequency has been further increased in the Covid crisis,” Reitbauer argues. In the Co-Innovation Lab at the Linz Institute of Technology (LIT), the spiral of complexity and growing data volumes is becoming a central research topic. Since one is not involved in short- and medium-term product development, one can collaborate intensively with the academic research community without economic pressure, argues Dynatrace Lab Director Andreas Hametner. “We can pursue and support exciting research approaches here at the Co-Innovation Lab, whose potential cannot yet be estimated.”

Univ. Prof. Rick Rabiser, head of the LIT Cyber-Physical Systems Lab and Hametner’s university counterpart at the Co-Innovation Lab, appreciates “the opportunity created by the Co-Innovation Lab to establish a new kind of transdisciplinary, scientific research between industry and university. Both partners, JKU and Dynatrace, bring different strengths to the cooperation. University research, motivated by practical challenges, has been a successful model at JKU for many years, especially in the field of software engineering. The Co-Innovation Lab enables JKU to further expand these strengths and to evaluate research methods based on real data. Dynatrace benefits from the closer connection to the academic research landscape, which can both create scientific foundations and provide the impetus for innovative solutions.”

Reinvention as a standing order

To drive profound innovation, Dynatrace retired a team of top developers from day-to-day operations back in 2014. Constant reinvention, as well as fully automated, self-healing, and self-protecting software, was the answer to growing complexity, he said. The artificial intelligence (AI) and machine learning required for this are the focus of research at the Co-Innovation Lab. To achieve this, new models must be developed for real-time data analysis (real-time analytics), for example. Since websites sometimes consist of several hundred microservices that constantly transmit data, they are confronted with rapidly increasing data volumes. “For this, we need to develop more efficient learning methods for AI to keep the multitude of models in real-time reporting up to date and thus also provide predictions of likely scenarios,” Hametner explains.

Alert inflation and learning models

Efficient training is also needed for the models used in data science. They must be able to recognize patterns in deeply individualized applications such as smartphone apps in order to make programs more stable and enable self-protection and self-healing. When it comes to security, there is another critical component to the complexity of the systems and the flood of data, Hametner explains. “Unfortunately, you have to assume that systems that were secure at release will not remain secure from attack. Security is the discipline with the most momentum.” That’s why it’s far from enough to identify problems early and report them reliably, he says. What matters is prioritizing between warnings of acute threats and, for example, minor problems in program libraries.

We cannot afford to overlook brainwaves was originally published in Dynatrace Engineering on Medium, where people are continuing the conversation by highlighting and responding to this story.