National Nuclear Laboratory


Sunday 25 May 2014

High Performance Cluster helps customers

Nexia Solutions has taken delivery of a new high performance computing cluster to support the modelling sector of our business. The improve hardware will ensure we continue to meet the increasing demands of our customers.

The new hardware, developed with IBM and OCF plc, will deliver over ten times thecombined performance of the legacy systems it will replace over the coming months.The cluster will form the core of a new modelling facility that is getting in shapefor the demands placed on Nexia Solutions and the forthcoming National NuclearLaboratory.Computing is one of the fastest changing areas of technology. So, the new clusterhas been designed from the outset using modular off-the-shelf components. Itprovides a range of processors and hardware configurations matched to the analysistypes needed, but within a centrally managed system. This means the cluster canbe expanded and refreshed rapidly to meet business needs. Despite deliveringunprecedented levels of computing power, the new cluster costs less than a quarter ofthe systems it is replacing.A dedicated high speed network called ‘infiniband’ enables individual CPUs to worktogether towards solving a single large model. We have already demonstrated thesolution of a complex fluid model shared between 128 processors. This solved in justa few seconds in comparison to several hours on a conventional PC.All of this computing power takes considerable resources to operate. Nexia Solutionshas partnered with ‘Digica’ to provide a dedicated purpose built hosting solution.Digica services provide the 100KW necessary to power and cool the system and alsoprovide UPS and generator power supply to ensure all-year-round service availability.This new level of performance will give Nexia Solutions the ability to deliver resultsfrom standard models much quicker. This will particularly benefit safety and criticalityteams, who have a requirement to respond at speed to safety related requests. Thiswill also benefit the Engineering Simulation Team who will be able to deliver resultsfrom models far quicker reducing delivery times and increasing value for money toour customers. These models include the Waste Vitrification Plant (WVP) melters atSellafield.The new cluster will open up some exciting opportunities for innovation. Projectsuggestions received so far include:• Application of advanced quantum mechanical techniques for probing the interactionof contaminates to metal surfaces• Modelling integrated processes, such as equipment subject to stress, corrosion andcracking• Coupling environmental and chemical models to give a detailed picture of risks ofradioactive contamination.The hardware will initially be available to modellers in the Safety and Criticality,Reactor Physics, Chemical, Engineering and Environmental teams. However, it isplanned that the service will be made available to modellers across the organisation inthe coming months.The new equipment will be fully integrated with the standard IT infrastructure networkand will be accessible from any Nexia Solutions site using web-based tools. Severalnew graphical workstations, to be located at Risley, Sellafield and Springfields, willprovide interactive model runs and post processing of results. The system is currentlyundergoing an extensive period of testing and will become fully operational before theend of 2007.Cluster To Muster New Computer Era

High Performance Computer Cluster – Improved Modelling Support – Scenario Examples• Engineering SimulationThe new system will enable us to respond to customer needs faster when running threedimensionalmodels of complex processing equipment, systems and chemical conditions,which evolve over time throughout operational lifetimes and in non standard emergencysituations. We will also be able to research into the modelling of integrated processes, whichwere previously modelled discretely. For example, modelling equipment integrity accountingfor electro-magnetic, computation fluid flow, thermal and mechanical stress analyses.Supplied by: Richard Field, Senior Technology Manager, Modelling and Simulation• Chemical ModellingThe new cluster will support the use of the latest quantum mechanical methods for chemicalproperty prediction from first principles. For example, Nexia Solutions can deploy the CASTEPcode, supplied by Accelrys Inc., to predict the spatial distributions, speciation and stability ofradioactive contaminates held within contaminated materials such as steels and concretes.The information gained by performing these simulations can be used to acceleratedecommissioning programmes by providing the information required to deploy the mostappropriate methods to remove contamination. For the first time, the cluster brings thepotential to deploy parallel processing to the solution of these models leading to benefitsvia delivery of calculations within realistic project timescales. This enables more routinedeployment of these state-of-the-art techniques to answer some of the more challengingquestions posed by customers.The availability of the new cluster will give Nexia Solutions access to some of the latestmolecular modelling techniques. For example, we can now deploy codes such as the LAMMPScode developed by Sandia National Laboratories for molecular and mesocale simulation ofmaterials behaviour. These codes take advantage of massive parallel processing, utilisinghundreds of individual CPUs to accelerate the solution of a single model. This brings benefitsof more realistic simulations designed to underpin and optimise chemical processes in areassuch as waste immobilisation, waste reprocessing and effluent treatment.Supplied by: Mark Bankhead, Research Technologist, Modelling and Simulation• Criticality and Safety AssessmentThe new cluster system brings many benefits to Criticality and Safety Assessment (NuclearSafety Section). Criticality models will be processed at much faster speeds, allowing forfar more detailed calculations to be conducted when compared to previous systems. Inaddition, the visualisation packages employed will be able to process the 2D and 3D imagesof the system being assessed more efficiently.Furthermore, the cluster will support an expanding team of criticality assessors, whichpreviously had to prioritise running of models due to the limited number of processors.The use of the new cluster will cover a wide range of criticality support activities suchas fuel production plants, waste management, decommissioning, residue recovery, fueltransportation, commissioning activities (including support to the South African Pebble BedModular Reactor project), validation and verification work as well as emergency/incidentresponse.These activities span a range of customers and, given the nature of the work, the resourcerequirements can peak extremely quickly over short or extended periods. This is anotherarea where the cluster will provide enhanced performance. The cluster management groupcan reallocate resources dependent on the need of any user group which would ensurethat customer requests for work can be addressed.There will also be enhanced local visibility of the integrity of the cluster system, with anyissues identified and addressed promptly.Supplied by: Andrew Brown, Criticality Team Managertalkingsolutions 08Facts and Figures• Powerful High Performance Computercluster running the latest Linux andWindows operating systems.• Equipped with:- 244×86 64-bit CPUs- 662GB memory- >10TB of dedicated fastnetworked attached storage• Accessible across sites via dedicatedsoftware applications and user-friendlyweb interface• Optimised to solve a diverse rangeof demanding computationalproblems in commercially driventimescales – examples:- Monte Carlo simulations tosolve safety issues relating to nuclearcriticality and R&D- Environmental and processengineering models coupling chemicaland transport properties- Predictive chemical modellingutilising state-of-the-art molecularand meso scale modelling techniques- Complex analytical simulations usingleading engineering modelling toolsincluding ‘Ansys’, ‘Dyan’ and ‘Fluent’