SCIENCE
An exclusive interview with Graeme Hackland, IT Manager, Renault F1 Team
- Written by: Chris O'Neal
- Category: SCIENCE
Supercomputing Online Live from SC08:At the SC08 show, Appro announced the final deployment of 38 TFLOP/s, Appro Xtreme-X Supercomputer for the new ING Renault F1 Team Computational Aerodynamics Research Centre. The supercomputer gives the research center more computing power to run full-car aerodynamic simulations using CFD applications required for improving the Renault F1 Team's racing cars. CFD is the science that their engineers apply to calculate the different flow of air around a moving car with the help of a SUPERCOMPUTER. To learn more, Chris O'Neal sat down with Graeme Hackland, IT Manager, Renault F1 Team for an overview of their new CFD supercomputer.
SC Online: What were the key challenges for adding new infrastructure at the ING Renault F1 Team? What were the specific business drivers that made you pursue a new solution like this one? Hackland: One of the most significant drivers was the fact that the alternatives were less cost effective. Combining this with the fact that new rules were placing more emphasis on CFD techniques and that both Renault F1 Team and Renault corporate could benefit from a new research centre made a strong business case for building the Computational Aerodynamics Research Centre. Given that the Supercomputer was to be housed in a new building integration with our infrastructure was simpler and less disruptive to the user community. The biggest challenge was the network integration. SC Online: Why did you choose Appro and why the Xtreme-X Supercomputer? Hackland: Appro not only offered us a cost effective solution but they also improved our required technical specification. The Xtreme-X Supercomputer gave us better reliability, greater fault tolerance and redundancy as well as more flexibility with regards to system scalability. This was all about the fact that Appro offered us a diskless, dual rail solution.
SC Online: Please describe your experience with Appro’s clusters and what is the performance, stability, scalability value for your applications? Also, please explain the benefits and your experience with the ACE- Appro Cluster Engine management system and its capability. How does it help simplify the work for the system administrators? Hackland: Mistral is ING Renault F1 Team’s first large scale Appro cluster deployment. The stability of the system is very good for all workloads. The scalability of the system allows capacity simulations to be run using a factor of 4 more processing cores then was previously possible with the current applications maintaining the same level of performance. Turnaround time of these simulations has been reduced by a factor 3 to allow more design iterations to be performed within the same time scales of the engineering design cycle. The system also allows much larger capability simulations for full scale car simulations with more complex physical models than was previously possible improving the accuracy of the simulations. The ACE software allows the team administrators to quickly and easily administer the system with minimum of effort while keeping the general management of the system simple, the ability to modify images and deploy within minutes across the whole cluster is especially useful SC Online: Please describe how you are using Appro's clusters and what specific projects and simulations are you using the system for? Hackland: Mistral is being used extensively for external aerodynamic development of next year’s car. Almost all areas of the car are being developed using CFD in conjunction with the wind tunnel, such as the front wing, diffuser, bodywork etc... Mistral is also being used for thermal analysis, for example to optimize the brake cooling. It represents a massive step forward for RF1 in terms of computing power allowing us to not only to run more simulations but to also run larger models and hence improve the accuracy of the simulations. The 2009 FIA regulations are very different from 2008; the increase in CFD throughput is allowing us to exploit the most in the shortest time from this regulation change. SC Online: Please describe the benefits that the Appro Xtreme-X Supercomputer brought to the ING Renault F1 Team based on the CPUs and interconnects. Hackland: The processor performance and Infiniband interconnect provide a balanced system for inter process communication and deliver the memory bandwidth and latencies required by the CFD applications used by Renault F1 Team. The Quad-core processors minimise the system footprint to which allows for future expansion of the cluster within the constraints of the new infrastructure. The processors also meet the power-performance requirements to minimise the environmental footprint of the system especially important to RF1 and inline with the environmental consideration of the new Computational Aerodynamics Research Centre. Supercomputing Online wishes to thank Graeme Hackland for his time and insight.
SC Online: What were the key challenges for adding new infrastructure at the ING Renault F1 Team? What were the specific business drivers that made you pursue a new solution like this one? Hackland: One of the most significant drivers was the fact that the alternatives were less cost effective. Combining this with the fact that new rules were placing more emphasis on CFD techniques and that both Renault F1 Team and Renault corporate could benefit from a new research centre made a strong business case for building the Computational Aerodynamics Research Centre. Given that the Supercomputer was to be housed in a new building integration with our infrastructure was simpler and less disruptive to the user community. The biggest challenge was the network integration. SC Online: Why did you choose Appro and why the Xtreme-X Supercomputer? Hackland: Appro not only offered us a cost effective solution but they also improved our required technical specification. The Xtreme-X Supercomputer gave us better reliability, greater fault tolerance and redundancy as well as more flexibility with regards to system scalability. This was all about the fact that Appro offered us a diskless, dual rail solution.
SC Online: Please describe your experience with Appro’s clusters and what is the performance, stability, scalability value for your applications? Also, please explain the benefits and your experience with the ACE- Appro Cluster Engine management system and its capability. How does it help simplify the work for the system administrators? Hackland: Mistral is ING Renault F1 Team’s first large scale Appro cluster deployment. The stability of the system is very good for all workloads. The scalability of the system allows capacity simulations to be run using a factor of 4 more processing cores then was previously possible with the current applications maintaining the same level of performance. Turnaround time of these simulations has been reduced by a factor 3 to allow more design iterations to be performed within the same time scales of the engineering design cycle. The system also allows much larger capability simulations for full scale car simulations with more complex physical models than was previously possible improving the accuracy of the simulations. The ACE software allows the team administrators to quickly and easily administer the system with minimum of effort while keeping the general management of the system simple, the ability to modify images and deploy within minutes across the whole cluster is especially useful SC Online: Please describe how you are using Appro's clusters and what specific projects and simulations are you using the system for? Hackland: Mistral is being used extensively for external aerodynamic development of next year’s car. Almost all areas of the car are being developed using CFD in conjunction with the wind tunnel, such as the front wing, diffuser, bodywork etc... Mistral is also being used for thermal analysis, for example to optimize the brake cooling. It represents a massive step forward for RF1 in terms of computing power allowing us to not only to run more simulations but to also run larger models and hence improve the accuracy of the simulations. The 2009 FIA regulations are very different from 2008; the increase in CFD throughput is allowing us to exploit the most in the shortest time from this regulation change. SC Online: Please describe the benefits that the Appro Xtreme-X Supercomputer brought to the ING Renault F1 Team based on the CPUs and interconnects. Hackland: The processor performance and Infiniband interconnect provide a balanced system for inter process communication and deliver the memory bandwidth and latencies required by the CFD applications used by Renault F1 Team. The Quad-core processors minimise the system footprint to which allows for future expansion of the cluster within the constraints of the new infrastructure. The processors also meet the power-performance requirements to minimise the environmental footprint of the system especially important to RF1 and inline with the environmental consideration of the new Computational Aerodynamics Research Centre. Supercomputing Online wishes to thank Graeme Hackland for his time and insight.