High-Performance Computing
Numerical computing and large quantities of data

HPC encompasses a rather broad range of technologies and common fundamental design considerations. Compute, memory, storage and communication bandwidth as well as delay budgets must be managed on top of highly individualized topologies. The questions to be answered computationally need to be broken down i.e. de-composited accordingly. Compute is anything but homogenous with various types of accelerators such as GPUs, FPGAs and other specialized types of hardware deployed in the field. HPC remains as one of the most technological challenging fields in information technology through all of its subdisciplines ranging from software development to hardware architecture. pleiszenburg.de offers near two decades worth of experience in the field, having been involved in the development and optimization of numerous highly customized scientific and engineering workloads.

Noteworthy interests and experience

  • Optimization of existing code for speed and/or memory consumption through various regimes
  • Fault-tolerant parallel computations
  • Up-scaling / re-adapting code and / or algorithms for parallel computations
  • supporting technologies: OpenMP, MPI, SIMD, BLAS, LAPACK, Dask, ZeroMQ, Apache Kafka, Slurm, Ansible, Salt, KVM, VirtualBox, Docker, Podman, VMWare, Hyper-V, Kubernetes, GlusterFS, Ceph, ZFS, iSCSI, Fibre Channel, PostgreSQL, PostGIS
  • Relevant languages: Python, C, C++, Rust, Fortran, Matlab, Basic, JavaScript, Bash
  • Design of processing pipelines based on Directed Acyclic Graphs (DAGs)
  • CUDA, starting with Compute Capability 3 (Kepler micro architecture) and CUDA Toolkit 4, via bare C/C++, numba (LLVM), pycuda or wrappers – depending on use-case
  • GPGPU with alternatives to CUDA, e.g.: openCL, ROCr / ROCm, Vulkan Compute, etc.
  • Designing and operating clusters backed by various types of Ethernet and/or Infiniband and storage
  • Scientific Data Storage: emphasis on numerical data and relevant technologies
  • Cloud: AWS, Azure, Google Cloud, Hetzner, Huawei and white label clones (as found e.g. in the middle east, among other regions), etc.
  • Deep knowledge of x86 architectures and CPUs
  • Deep knowledge of relevant linux operating system / kernel internals