Cardiovascular diseases are one of the most critical health issues in developed countries, affecting 30% of the population. The advances in modern medicine have been mainly based on experimental research but during the last decade, in-silico experimentation has provided a powerful tool to understand physiology in health and disease and to design surgical procedures or devices.

PASIM takes profit of the extensive research on computational biomechanics done at Barcelona Supercomputing Center by transferring this technology to the users that can take the most advantage of it. Through the democratisation of supercomputing resources using cloud-based technology and reliable simulation results validated following the most strict ASME V&V40 standard, our goal is to put this tool to the device designers and clinicians fingertips.

PASIM first milestone is to provide a regulatory numerical test to augment the data provided during pre-market approval, therefore reducing time-to-market of novel Left Ventricular Assist Devices (LVADs). LVADs are implantable blood pumps that complement heart function on severe heart failure, generally used as a bridge to heart transplant. The scarce animal test done during pre-market evaluation of the LVAD tend to misrepresent the blood damage produced by the pump on human clinical usage. This regulatory numerical test allows to better predict the clot formation o a determined LVAD model, reducing patient risk and time-to-market.

We found valuable the entire set of activities composing the BPP program. During the BPP course in Goteborg we learn how to see from a different angle the research products by following the business canvas method. During the workshop at HiPEAC we had the opportunity to receive feedback and steering on how to tune the language of a pitch and the way to create a roadmap toward the catch of a business perspective. Such gained experience has been used to present in February (right before the lockdown) at the cybersecurity center of the New York 12 University, the entire GreenHLS toolset with the goal to create a path where our product can start to be perceived as a “brick” for other possible exploitations. At the same time, we realized which are the possible showstoppers to the use of our methodology from a wider public and we are also working with our Technology Transfer Office to understand how we can distributed at least a free version of the tools, to gain acceptance and create awareness, while maintaining the control over the developed IPs. We believe that this is a crucial point before to proceed with other more related marketing steps. As a general suggestion, the aspects related to the licensing and IP protection can be a topic to be covered with more extent during the next editions of the BPP program.

Background and Overview

Ongoing efforts to digitize activities and processes across the public sector and within commercial organisations and across entire industry sectors are driving the development of optimized use-cases for High Performance Data Analytics (HPDA). HPDA directly enables efficient business intelligence, and hence value to be extracted from the data produced by an organization’s operations. HPDA solutions are often constructed with the aid of Apache JVM-based frameworks (such as Flink, Spark and Beam) and deployed on HPC bare-metal or cloud resources using containers and virtual machines with access to persistent cloud storage frameworks, such as Amazon’s S3.
Significant barriers to the implementation and efficient exploitation of HPDA digitization activities concern the complexities of performance optimization, configuration for deployment, capacity planning, and also the relatively small and expensive talent pool of personnel with appropriate technical experience spanning application frameworks and cloud technologies. Customers and organizations seeking to exploit HPDA must therefore incur the high costs of building their own internal teams with appropriate expertise, and/or, enter into consultancy/support contracts with individuals and companies that focus on providing such services.
The PErformance Tuning Deployment Optimisation for High Performance Data Analytics (PETO4HPDA) BPP project has sought to identify potential sustainable business opportunities that can leverage the University of Manchester’s BCC/eBPF https://github.com/iovisor/bcc and MAMBO https://doi.org/10.1145/2896451 expertise in Linux tracing and instrumentation, with the aim of providing customised performance monitoring and analysis tools that can enable efficient exploitation of HPDA frameworks. Market research interviews were conducted to investigate the business needs of a wide range of potential customers, and to identify customer value propositions that were not adequately met by current tools and techniques.

Outcomes

The results of the project have been captured as a lean business model canvas that has identified promising potential customer value propositions, and indicative venture capital investment requirements for proceeding to a startup company. PETO4HPDA has identified and is currently investigating how its technological innovation may benefit performance optimisation of representative benchmarks from a potential early adopter customer involved in B2B support for e-commerce. Prototype tool functionality has primarily focussed on BCC/eBPF monitoring and analysis, and the current goal is to be able to meet the performance optimisation needs of JVM-based products implementing B2B support for e-commerce related to the early adopter. The project has identified that BCC/eBPF technology may offer suitable solutions at lower over overhead than current techniques for security, JavaScript/NodeJS and Android that were beyond the scope of PETO4PHA’s initial remit focussing on HPDA. A white paper is to be produced to showcase the capabilities of prototype BCC/eBPF tools for analysing Java application performance using fine grain tracing of performance counters and the occurrence of JVM events and operations of interest.

At the current moment in time, PETO4HPDA is not ready to progress for applying for startup funding, because the business model, links to multiple early adopters, and testing of prototype tools need further development and elaboration. Consequently, we have applied for, and been awarded further funds to continue development of PETO4HPDA under a follow up project funded by the University of Manchester’s Researcher 2 Innovation (R2I) scheme. The R2I scheme will further refine the business model canvas, by performing market research and interviews with the aim of further exploring HPDA, and additional potential business opportunities, within security and performance analysis for Android applications.