An Autonomic Performance Environment for Exascale

Abstract

Exascale systems will require new approaches to performance observation, analysis, and runtime decision-making to optimize for performance and efficiency. The standard "first-person" model, in which multiple operating system processes and threads observe themselves and record first-person performance profiles or traces for offline analysis, is not adequate to observe and capture interactions at shared resources in highly concurrent, dynamic systems. Further, it does not support mechanisms for runtime adaptation. Our approach, called APEX (Autonomic Performance Environment for eXascale), provides mechanisms for sharing information among the layers of the software stack, including hardware, operating and runtime systems, and application code, both new and legacy. The performance measurement components share information across layers, merging first-person data sets with information collected by third-person tools observing shared hardware and software states at node-and global-levels. Critically, APEX provides a policy engine designed to guide runtime adaptation mechanisms to make algorithmic changes, re-allocate resources, or change scheduling rules when appropriate conditions occur.

Harnessing billions of tasks for a scalable portable hydrodynamic simulation of the merger of two stars

“…The HPX implementation is also integrated with the Autonomous Performance Environment for eXascale (APEX) (Huck et al, 2015;XPRESS APEX, 2017). The APEX environment provides a lightweight performance measurement and control library specifically designed for use in HPX and has been extended to other runtime systems.…”

“…Parallel programming models are emerging to meet the challenges of manycore, heterogeneous, and exascale architectures. Models call on the AMT methodology to efficiently avoid artificial barriers and overlap communication with computation to hide unavoidable latencies (Anderson et al, 2013;Dekate et al, 2012;Huck et al, 2015;USDOE, 2012;Wheeler et al, 2008). While the concepts of AMT systems are emerging in many modern programming models, some of the more advanced competing implementations of new programming models with similar concepts include Charmþþ (Kumar et al, 2004), Intel Cilk Plus (Intel, 2017a), OpenMP with tasking (deSupinski et al, 2017), Chapel (Chamberlain et al, 2007), Intel SPMD Program Compiler (Intel, 2017b), X10 (Charles et al, 2005), and Legion (Bauer et al, 2012).…”

HPX - The C++ Standard Library for Parallelism and Concurrency

“…Policy Engine/Policies (Huck et al, 2015;Khatami, Troska, Kaiser, Ramanujam, & Serio, 2017;Laberge et al, 2019) Often, modern applications must adapt to runtime environments to ensure acceptable performance. Autonomic Performance Environment for Exascale (APEX) enables this flexibility by measuring HPX tasks, monitoring system utilization, and accepting user provided policies that are triggered by defined events.…”

A modelling framework for simulating large geographical systems of agents and fields

“…Having to determine this partition size is something we would like to relieve the modeller of, possibly by an automatic procedure. The integration of the APEX performance environment for runtime adaptation [81] would allow for the automatic selection of best partition sizes, for example.…”