The cancellous bone simulation use case deals with the need of hospitals to run simulations of cancellous bones. The use case provides two major advancements on the state of the art. First, we will reach far greater flexibility in executing and maintaining the simulations. Second, the use case will allow for interactive real-time simulations, which currently is not possible.
Cancellous bone simulations are a simulation technique that models the interaction of implants with cancellous bones. These simulations provide data that is taken into account in medical diagnosis regarding the positioning of implants for certain pathologies. Cancellous bone simulation is a special method that overcomes the shortcomings of similar, simplified simulation techniques. The goal of cancellous bone simulations is to provide accurate decision support to physicians and surgeons, which will lead to better fitting bone implants. The better fit, in turn, improves patients’ quality of life.
Cancellous bone simulations require a lot of computational resources and thus are dependent on an HPC system. However, typically hospitals cannot afford large HPC systems for specific simulations such as cancellous bone simulations. Thus, in this use case hospitals run their simulations on HPC systems from third parties, in this case HLRS. One downside of this setup is that the simulations need to be tailored to the specific HPC system. To achieve good performance the simulations need to fit the underlying hardware, which is costly. Furthermore, the platform dependence introduces a high threshold to migrate the simulation to other hardware. Another disadvantage, of the current HPC deployment, stems from the batch systems that HPC systems run. The batch systems make the interactive real-time use of cancellous bone simulations impossible. Real-time interaction, however, could further assist surgery, help to detect problems with implants, and improve the production process of implants.
In this use case, we will apply Mikelangelo’s technology stack for an HPC use case. The goal is to provide fast execution and high flexibility for cancellous bone simulations. Our I/O-optimized hypervisor will improve I/O performance, which currently poses a major barrier to run computations on virtual machines. To achieve further speedup, we will use OSv with paravirtualization for zero-copy RDMA. To deploy and manage the simulations, we will use the adapted cloud middleware.
To execute the cancellous bone simulation in a cloud, we need to adapt the simulation code itself. HLRS employs the original developers of the simulation, and it manages its own HPC system. Furthermore, HLRS has already ported the simulation for different machines in the past. Thus, HLRS is well qualified to port the simulation into the cloud. All dependencies for the cancellous bone simulation are available for Linux, which will make it possible to port the simulation to OSv, as well.
In this use case, we will port the cancellous bone simulation, onto an HPC-cloud and a commodity cloud. We will use the port to the HPC-cloud, to evaluate the performance of the in a virtualized setting with the bare-metal performance. Furthermore, we will compare these performance records to the performance in a general-purpose cloud, as it might be used by end-users.
We expect multiple beneficial results for various stakeholders of cancellous bone simulations. First, through increased flexibility, hospitals’ stale IT resources can be used to run simulations on-premise. The same flexibility reduces the maintenance cost for the simulation code. As an effect, hospitals can save money by running their simulations on-premise. Second, the increased flexibility allows hospitals to perform more simulations, with no added costs, by outsourcing the computation, as they do now. Third, interactive real-time simulations will become available to as a new usage scenario. Interactive simulations will enable clinicians to provide higher quality services to their patients.
Conclusively, this use case will leverage Mikelangelo’s stack to improve the quality of life for patients and reduce costs for hospitals running cancellous bone simulations.