SKA computing an HPC challenge

The exascale radio telescope Square Kilometre Array (SKA) [1] will require supercomputers with high technical demands. The Science Data Processor (SDP) pipeline in charge of producing the multidimen- sional images of the sky will have to execute in realtime a complex algorithm chain with data coming from telescopes at an incredible rate of several Tb/s and limited storage possibilities. The SDP will also have to be as green as possible with an energy budget of only 1 MWatt for 250 Petaflops.

The SDP supercomputer will be based on a standard HPC system combined with FPGA or application-specific architectures like GPU or the manycore Kalray Massively Parallel Processor Array (MPPA). One crucial challenge is to assess the performance both in time and energy of new complex scientific dataflow algorithms on not-yet-existing complex computing infrastructures. It will be hardly possible without efficient co-design methods and rapid prototyping tools.

Dark-Era Goals

  1. Building SimSDP, a rapid prototyping tool providing exascale simulations from dataflow algorithm description.
  2. Exploring low power accelerators like FPGA or Kalray MPPA as alternatives to mainstream GPU architecture.
  3. Being source of proposals for SKA computing and promoting french contributions such as ddfacet [4].

Dark-Era consortium

The consortium gathers complementary skills in computer science, signal processing, and astronomy with twelve permanent members from the SimGrid [2] development Team at IRISA, the PREESM [3] development team at IETR, the inverse problem team at L2S, and two radio astronomy teams at Observatories of Paris and Coˆte d’Azur. Dark-Era has the support of SKA-France and works in collaboration with Atos-Bull.

SimSDP, an exascale simution tool

SimSDP purpose is to provide early analyses in terms of mem- ory usage, latency, throughput, and energy consumption through an original mixed approach based on execution and simulation. Following an Algorithm Architecture Matching (AAM) approach, SimSDP will rely on a dataflow model of the algorithm and a model of the target architecture. SimSDP will be based on two existing tools: PREESM and SimGrid. PREESM accurately evaluates heterogeneous single node performance; SimGrid accurately simulates inter-node communications. Their association will allow for reliable simulations of large scale heterogeneous HPC systems.

  • [1] “SKA Organisation public website”, https://www.skatelescope.org
  • [2] H.Casanova et al., “Simulation of Distributed App.”, J. Parallel Distrib. Comput.
  • [3] “PREESM,an open source rapid prototyping tool”, https://preesm.github.io.
  • [4] C.Tasse et al.,“Faceting for direction-dependent spectral deconvolution”, A&A,2018.