Towards an Environment to Deliver High Performance Computing to Small and Medium Enterprises

Continuous Program Optimization via Advanced Dynamic Compilation Techniques

“…To verify this assertion we compiled and run the PolyBench/C benchmarks and we report in Figure 3 runtime and compilation time for each benchmark. The run 2 https://github.com/skeru/polybench_libVC time of the different code versions generated using libVC compilers is not significantly influenced by the chosen compiler alternative. On the contrary, this choice becomes relevant for the compilation time.…”

An ontology for heterogeneous resources management interoperability and HPC in the cloud

“…Other initiatives in the EU frameworks FP7 and H2020 targeting interoperability and lock-in vendor issues in cloud computing environments have been explored. Foremost among these are: Cloud4-SOA [39], a platform that performs seamless adaptive multi-cloud management, semantically interconnecting heterogeneous PaaS offerings across different cloud providers with the same technology and supported by Apache Brooklyn blueprints and TOSCA extensions; PaaSport [40] where Cloud PaaS technologies are combined with lightweight semantics and in which application models and SLAs follow a Descriptions and Situations Pattern Technique, used to deliver a thin and non-intrusive Cloudbroker in the form of a Cloud Marketplace; Paassage [41] that constructs a deployment mechanism for applications across public and private clouds constrained by a set of rules described in the CAMEL [42] modeling language; ModaClouds [43] that follows a Model-Driven-Development for Clouds and Multi-Clouds, performing semi-automatic translations into code to enable deployments across public and hybrid cloud vendor platforms; and finally, the Mikelangelo project [44], that provides support for HPC in Cloud environments by implementing a bespoke virtual machine manager to allow the management in of HPC and Cloud resources. Although these projects succeed in exploring different alternatives for interconnecting heterogeneous vendor service offerings, enhancing extant interfaces, or providing decision support systems, either they neglect the incorporation of HPC into the Cloud and the subsequent need for having multiple resource management domains, or they do not support a wider heterogeneous environment, being constrained to only resource management, hardware accelerators, or resource abstractions in the same Cloud.…”

The RECIPE approach to challenges in deeply heterogeneous high performance systems

“…The "REliable power and time-ConstraInts-aware Predictive management of heterogeneous Exascale systems" (RECIPE) project targets such emerging HPC scenarios, characterized by new types of requirements, including time criticality and reliability, as well as new delivery models, e.g. cloud-based solutions [2], potentially open to a large audience of HPC users. These emerging scenarios require that new HPC platforms are able to support multiple applications running concurrently, possibly with conflicting Quality of Service requirements [3][4][5][6][7].…”