Software speeds applications in Austin
Software for data analysis, system management, and for debugging other software will be among the innovations to be announced and on display at SC15, the US Supercomputing Conference and Exhibition, which is taking place in Austin Texas from 15 to 20 November. In addition to the software, novel and improved hardware will also be on display, together with an impressive array of initiatives from Europe in research and development leading up to Exascale computing.
Software tools for developers and system administrators are the focus of this, the first of four news reports on what will be on display in Austin. These update and supplement the information already available from the Show Preview.
Among the software tools for developers, Allinea Software will demonstrate significant extensions to its Forge integrated development tool suite as well as its Performance Reports analytics tool. Version 6.0 of both products will assist developers, users, analysts and system administrators not only on Intel Xeon and Xeon Phi processors, but also those interested in running on ARM 64-bit and OpenPower processors.
The Allinea Forge debugger, DDT, now has leaked pointer tracking, which is designed to solve particularly tricky memory bugs. DDT's non-interactive ‘off-line’ mode adds live feedback on the progress of an application and allows users to take ‘snap-shots’ of the state of the program during execution.
To enhance a code’s performance, a new source-line instruction analysis, MAP, in the Allinea Forge profiler will assist developers, by showing -- alongside the source code -- how long each line took and how it spent that time, whether in floating point, vector instructions, or memory accesses.
User-written custom metrics are also added to MAP – enabling users to add performance insight relevant to their own applications or domains – such as domain size or boundary element counts to track the impact of adaptive meshes on performance over time.
Also on the theme of helping developers to program and execute their applications efficiently, the Workflows and Distributed Computing team at the Barcelona Supercomputing Centre is releasing version 1.3 of the programming environment COMPSs, incorporating new features to improve runtime performance, provide better support for Python, and a new monitoring tool
COMPSs is a task-based programming model that automatically parallelises the execution of large scale applications, allowing them to run efficiently on distributed computational infrastructures such as clusters, grids, and clouds.
COMPSs has been available for the last years to users of the MareNostrum supercomputer and to the Spanish Supercomputing Network and has been adopted in research projects across disciplines as diverse as biomedicine, engineering, biodiversity, chemistry, astrophysics, and earth sciences. There is a new implementation of the workers whose execution time now persists during all the application lifetime, reducing runtime overhead. Python support has been extended with constraints support and support for user decorators. COMPSs is now offered with a new monitoring tool, capable of showing the progress of the application as well as details about resources’ usage.
COMPSs has recently attracted interest from areas such as image recognition, genomics and biodiversity. Efforts have also been focused on emerging virtualisation technologies, adopted by cloud environments. In such systems, COMPSs provides scalability and elasticity features by dynamically adapting the number of resources to the actual workload. COMPSs is interoperable with both public and private cloud providers such as Amazon EC2, OpenNebula, and with OCCI compliant offerings.
Bright Computing, which provides hardware-agnostic cluster and cloud management software, will release several updates and enhancements to its most popular management software solutions at SC15. The updates will include more than a dozen features that simplify OpenStack deployment, add built-in integration functionality, and greatly improve Big Data integration, including support for the latest releases from Apache, Cloudera, Hortonworks, and Pivotal.
‘Bright’s HPC cluster management solution provides a solid foundation for combining high-performance computing, big data, and private cloud environments,’ said Dr Matthijs van Leeuwen, founder and CEO of Bright Computing. ‘Our latest enhancements respond to the growing trend we are seeing in which more and more HPC users want to combine HPC workloads with big data analytics workloads in the same infrastructure.’
As users demand flexibility to run workloads on either bare metal, virtualised machines, or in containerized environments, Bright has been focusing on developing a complete infrastructure that gives operators this choice. ‘Our management solutions offer a “single pane of glass” management for the hardware, the operating system, the software, and users. Administrators can offload the workload to a public cloud or to a mixed environment, with some servers run on-premises and some in the cloud.’
eXcellence in IS Solutions (X-ISS) will unveil plans at SC15 to expand its signature ManagedHPC service into the European market. Also at the show, X-ISS will demonstrate new capabilities of its DecisionHPC business analytics software and provide insights into its recently announced CloudHPC consulting service.
ManagedHPC is a remote cluster administration and monitoring service that has maximised HPC performance for engineering, oil and gas, health science, and governmental organisations in the USA for more than 15 years. In response to international requests, X-ISS will now offer the service to European organisations with live personnel support during regular business hours across the continent.
Designed specifically for small- and medium-sized clusters in enterprises that lack the internal resources or expertise to manage HPC environments, the ManagedHPC service provides daily remote monitoring of the systems, keeping them running at their peak efficiency. ManagedHPC services are also retained by large organisations with clusters located at multiple facilities.
Deepak Khosla, president and founder of X-ISS, said: ‘ManagedHPC positively impacts cluster ROI by spotting problems before they become critical, thus reducing system downtime.’