The UberCloud Experiment Continues…

By Wolfgang Gentzsch and Burak Yenier

June 16, 2014

Improved product quality, faster time to market and ultimately increased ROI have long been associated with the adoption of HPC tools. The benefits that engineers and scientists can expect from using technical computing in their research, design and development processes can be huge. In spite of this, there is still unmet need with relatively few scientists and manufacturers using servers or clusters when designing and developing their products on computers. The majority of virtual prototyping and large-scale data modeling is still performed on desktop or laptop computers. It’s not surprising then that many of these users face problems stemming from the lack of performance of their machines. Obtaining additional computing power means either purchasing a server outright or using a cloud-based offering.

Many system vendors have developed a complete set of products, solutions and services for “high performance computing” (HPC) and buying an HPC server for an SME is no longer out of reach. Owning an HPC server, however, is not necessarily the best idea in terms of cost-efficiency or maintenance, because the Total Cost of Ownership (TCO) of such a machine is pretty high, and maintaining such a system requires additional manpower and expertise.

The other option today is to use an HPC Cloud solution that allows engineers and scientists to keep using their workstation for daily design and development work, and to “burst” larger, more complex jobs into the Cloud when needed. Thus, users have access to quasi infinite computing resources that offer higher-quality results. A Cloud solution helps reduce capital expenditure, offers greater business agility by dynamically scaling resources up and down as needed, and is only paid for when used.

The UberCloud Experiment Accelerating HPC Cloud Adoption

Since July 2012 the UberCloud Experiment has attracted more than 1500 organizations from 72 countries. We were able to build 152 teams in CFD, FEM, Biology, Finance, and other domains, and to publish many case studies reporting about different applications, experience, and lessons learned. UberCloud TechTalk has been founded providing educational lectures for our community. And the UberCloud Exhibit offers a Cloud services catalogue where community members can exhibit their Cloud related services or select the services which they want to use for their team experiment or for their daily work. Sponsored by Intel and HPCwire in 2013, the first Compendium with 25 HPC Cloud case studies received the HPCwire Readers Choice Award for the best HPC Cloud implementation. Now, in June 2014, the second Compendium of UberCloud case studies has been published, sponsored by Intel and HPCwire, which can be downloaded for free from here.

The UberCloud Experiment provides a platform for scientists and engineers to explore, learn and understand the end-to-end process of accessing and using HPC Cloud resources, and to identify and resolve the roadblocks. End-users, software providers, resource providers, and computing experts collaborate in teams to jointly solve the end-users’ application in the Cloud. Let’s start by defining what roles each stakeholder plays to make service-based high performance technical computing in the cloud come together:

End-user
– A typical example is a small or medium size manufacturer in the process of designing, prototyping and developing its next-generation product. These users are prime candidates for HPC-as-a-Service when in-house computation on workstations has become too lengthy and acquiring additional computing power in the form of a server is too cumbersome.

Application software provider
– These are software owners of all stripes, including ISVs, public domain software organizations and individual developers. The UberCloud Experiment prefers rock-solid software, which has the potential to be used on a wider scale.

Resource provider
– This pertains to anyone who owns HPC resources networked to the outside world. A classic HPC center would fall into this category as well as a standard datacenter used to handle batch jobs, or a cluster-owning commercial entity that is willing to offer up cycles to run non-competitive workloads during periods of low CPU-utilization.

Computing experts
– This group includes individuals and companies with HPC expertise, e.g. in areas like cluster management and software porting and optimizing. It also encompasses PhD-level domain specialists with in-depth application knowledge. These experts work as team leaders, with end-users, computer centers and software providers to help glue the pieces together.

For example, suppose the end-user is in need of additional compute resources to increase the quality of a product design or to speed up a product design cycle – say for simulating more sophisticated geometries or physics or for running many more simulations for a higher quality result. That suggests a specific software stack, domain expertise and even hardware configuration. The general idea is to look at the end-user’s tasks and software and select the appropriate resources and expertise that match certain requirements.

Cloud Case Studies from the UberCloud Experiment

As a glimpse into the wealth of practical use cases, we chose four HPC Cloud projects out of the 152 UberCloud Experiments: in life sciences teams 61 and 89, and in CAE teams 118 and 142.

Team 61: Molecular dynamics of the mutant PI3Kα protein

The end-user of this team was Zoe Cournia from the Biomedical Research Foundation of the Academy of Athens. Resource provider was the GRNET-Okeanos IaaS Cloud service, represented by Vangelis Floros and Stefanos Gerangelos from the Greek Research and Technology Network S.A. And HPC expert was Dominique Dehareng from the Center for Protein Engineering at the University of Liège. Open source software Gromacs 4.6.1 was used for the simulations.

Cancer is a leading cause of death worldwide, accounting for 7.6 million deaths in 2008 according to the World Health Organization. This number is expected to increase to 21 million by 2030. One of the signaling pathways which, when deregulated, becomes centrally involved in several types of cancers, like colon, mammary, and endometrial tumorigenesis, is served by phosphoinositide-3-kinase alpha (PI3Kα). The importance of PI3Kα in cancer is highlighted by the discovery that PIK3CA, the gene encoding the catalytic p110α subunit of PI3Kα, is frequently mutated in human malignancies. The goal of this project was to gain insights into the oncogenic mechanism of two commonly expressed PI3Kα mutants by studying their conformational changes with Molecular Dynamics (MD) simulations, in comparison with the PI3Kα wild-type (normal, non-cancerous) protein. The utilization of cloud computing in performing MD simulations of mutant PI3Kα with Gromacs was examined in this case study.

UberCloud_Fig1

Figure 1: The protein PI3Ka is depicted in ribbons and is placed in a water box, shown as red dots.

The cloud computing service that was provided was a free and very efficient service. Cloud computing proved to be an extremely easy process as compared to building and maintaining your own cluster. The team was provided with a VM with 8 cores and the possibility of building a cluster of up to 64 cores connected via Ethernet network. In the 8 cores, the MD simulation of the mutant PI3Kα protein scaled linearly and did run faster when compared to in-house simulations. Team communications through BaseCamp were very efficient. Collaborative work over the Internet, using on-line resources like cloud computing hardware and open source software such as Gromacs, is an efficient alternative to in-person meetings and in-house calculation servers.

More details about this case study can be found in the second UberCloud Compendium which can be downloaded here for free.

Team 89:
Performance Analysis of GROMACS Molecular Dynamics for Simulating Enzyme Substrate in the Cloud

The end-users of this team were Pravin Kumar R, Thirupathi Jatti, Soma Ghosh, Satish Nandagond, and Naveen Kulkarni from Polyclone Bioservices. Patrice Calegari and Marc Levrier were from resource provider Bull ServiWare providing eXtreme factory cloud services. Jaap Flohil from Foldyne was the team expert and Dennis Nagy from BeyondCAE the UberCloud mentor of the team.

The team focused on evaluating the performance of double precision MPI-enabled GROMACS 4.6.3 on 25 Bullx 510 blades (each 16-core Intel SNB, total 400 cores) for peta-scaling molecular dynamics simulations. The activities were organized into three tasks: install and optimize GROMACS performance on the Bull extreme factory cluster; install accessory tools to analyze simulation data; and test different multi-scale molecular systems involving enzyme substrate complexes on the cluster. The starting point of eEF (enzyme engineering framework), Polyclones’ framework for enzyme engineering, is to conduct molecular dynamics (MD) studies and calculate different parameters using the MD trajectories.

UberCloud_Fig2

Figure 2: Protein enclosed in a box of water and ions. We can study proteins in atomic detail, down to the movements of individual water molecules (red/white balls and stick model), ions (purple), and the protein itself shown in the surface model (blue and green represents macromolecular dimer form of the protein).

Using 400 cores for MD studies was extremely helpful in estimating the time and resource for the partial completion of an enzyme engineering project using eEF. Bandwidth and latency were excellent. The remote visualization system was used to analyze huge (100 GB) simulation data. The web interface from the eXtreme factory cloud helped in shooting the jobs, allocating the cores for different jobs simultaneously, and also organizing the output results. The XF team installed and built a web interface for Gromacs, which can be used seamlessly to run and analyze molecular dynamics.

UberCloud_Fig3

Figure 3: Performance trend as the number of cores increases for a system with ~120K particles.

More details about this case study can be found in the second UberCloud Compendium which can be downloaded here for free.

Team 118:
Coupling In-house FE Code with ANSYS Fluent CFD

The end user was Hubert Dengg from Rolls-Royce Deutschland, software providers were Wim Slagter and René Kapa from ANSYS, resource providers and team experts were Thomas Gropp and Alexander Heine from CPU 24/7, and Marius Swoboda from Rolls-Royce Deutschland acted as HPC/CAE expert.

In the present test case, a jet engine high pressure compressor assembly was the subject of a transient aerothermal analysis using FEA/CFD coupling technique. Coupling is achieved through an iterative loop with smooth exchange of information between the FEA and CFD simulations at each time step, ensuring consistency of temperature and heat flux on the coupled interfaces between the metal and the fluid domains. The aim of the HPC Experiment was to link ANSYS Fluent with an in-house FEA code. This was done by extracting heat flux profiles from the Fluent CFD model and applying them to the FE model. The FE model provides metal temperatures in the solid domain.

This conjugate heat transfer process is very consuming in terms of computing power, especially when 3D CFD models with more than 10 million cells are required. As a consequence, we thought that using cloud resources would have a beneficial effect regarding computing time.

The computation was performed on the 32 cores of two nodes with dual Intel Xeon processors. The calculation was done in cycles in which the FE code and Fluent CFD ran alternating, exchanging their results.

UberCloud_Fig4

Figure 4: Contours of total temperature for a jet engine component.

Outsourcing of the computational workload to an external cluster allowed the end user to distribute computing power in an efficient way – especially when the in-house computing resources were already at their limit.

Bigger models usually give more detailed insights into the physical behavior of the system.

In addition, the end user benefited from the HPC provider’s knowledge of how to setup a cluster, run applications in parallel based on MPI, create a host file, handle licenses, and prepare everything needed for turn-key access to the cluster.

More details about this case study can be found in the second UberCloud Compendium which can be downloaded here for free.

Team 142:
Virtual Testing of Severe Service Control Valve

End User was Mark A. Lobo from Lobo Engineering.
Autodesk provided Simulation CFD 360 (SimCFD) and the supporting cloud infrastructure. And the HPC/CAE application experts were Jon den Hartog and Heath Houghton from Autodesk.

UberCloud_Fig5

Figure 5: a control valve model with idealized flow path was used to minimize effects of a complex body cavity and trim design.

Flow control valve specifications include performance ratings in order for a valve to be properly applied in fluid management systems. Control systems sort out input parameters, disturbances and specifications of each piping system component to react and produce a desired output. System response is chiefly a function of the accuracy of control valves that respond to signals from the control system. Valve performance ratings provide information to the system designer that can be used to optimize control system response.

The premise of this project was not only to explore virtual valve testing, but to evaluate the practical and efficient use of CFD by the non-specialist design engineer. As a benchmark, the end user had no prior experience with the Autodesk software when the project initiated, no formal training in the software, and he was depended on the included tutorials, help utility, thorough documentation to produce good results and good data.

UberCloud_Fig6

Figure 6: Application Domain. The control valve restriction components or “trim” reduces the annular area as the cavity profile on the right moves to the left. The location of highest velocity is indicated in red.

One of the benefits for the end-user was that cloud computing enabled accessing a large amount of computing power in a cost effective way. Rather than owning the hardware and software licenses, engineers can pay for what they need when they need it, rather than making a substantial upfront investment. In this project, over 200 simulations were run in the cloud. Given the runtimes involved and allowing for data download upon completion of the runs, it is possible for all of these simulations to be solved within a day. For an engineer with 1 simulation license on a single workstation, this would have required 800 hours (approximately 30 days) to complete if the simulations were running nonstop one after another. Table 1 compares the approximate time and investment that would be required for various solving approaches.

Table : Comparison of Desktop, Cloud, and HPC Solving Options

Simulation Solving Approach

Approx Time to Complete

Investment Required

Local Desktop Machine

800 hours

(1 month)

Engineering Workstation +Simulation SW License
Local Desktop Machine + Cloud Computing

24 hours

(1 day)

Engineering Workstation +Simulation SW License +

$1200 Cloud Compute Fee

Local Desktop Machine + Private HPC Cluster + Multiple Solver Licenses

24 hours

(1 day)

Engineering Workstation +Simulation SW License +

30 Node Compute Cluster +

30 Simulation Solver Licenses

 

More details about this case study can be found in the second UberCloud Compendium which can be downloaded here for free.

The 2nd UberCloud Compendium with HPC Cloud Case Studies is intended as a resource for engineers, scientists, managers and executives who believe in the strategic importance of HPC as a Service, in the Cloud. It’s a collection of selected HPC Cloud case studies from the participants in Rounds 3 & 4 of the UberCloud Experiment. Among these case studies are candid descriptions of challenges encountered, problems solved, lessons learned, and recommendations. This second UberCloud Compendium can be downloaded here.

Subscribe to HPCwire's Weekly Update!

Be the most informed person in the room! Stay ahead of the tech trends with industry updates delivered to you every week!

MLPerf Inference 4.0 Results Showcase GenAI; Nvidia Still Dominates

March 28, 2024

There were no startling surprises in the latest MLPerf Inference benchmark (4.0) results released yesterday. Two new workloads — Llama 2 and Stable Diffusion XL — were added to the benchmark suite as MLPerf continues Read more…

Q&A with Nvidia’s Chief of DGX Systems on the DGX-GB200 Rack-scale System

March 27, 2024

Pictures of Nvidia's new flagship mega-server, the DGX GB200, on the GTC show floor got favorable reactions on social media for the sheer amount of computing power it brings to artificial intelligence.  Nvidia's DGX Read more…

Call for Participation in Workshop on Potential NSF CISE Quantum Initiative

March 26, 2024

Editor’s Note: Next month there will be a workshop to discuss what a quantum initiative led by NSF’s Computer, Information Science and Engineering (CISE) directorate could entail. The details are posted below in a Ca Read more…

Waseda U. Researchers Reports New Quantum Algorithm for Speeding Optimization

March 25, 2024

Optimization problems cover a wide range of applications and are often cited as good candidates for quantum computing. However, the execution time for constrained combinatorial optimization applications on quantum device Read more…

NVLink: Faster Interconnects and Switches to Help Relieve Data Bottlenecks

March 25, 2024

Nvidia’s new Blackwell architecture may have stolen the show this week at the GPU Technology Conference in San Jose, California. But an emerging bottleneck at the network layer threatens to make bigger and brawnier pro Read more…

Who is David Blackwell?

March 22, 2024

During GTC24, co-founder and president of NVIDIA Jensen Huang unveiled the Blackwell GPU. This GPU itself is heavily optimized for AI work, boasting 192GB of HBM3E memory as well as the the ability to train 1 trillion pa Read more…

MLPerf Inference 4.0 Results Showcase GenAI; Nvidia Still Dominates

March 28, 2024

There were no startling surprises in the latest MLPerf Inference benchmark (4.0) results released yesterday. Two new workloads — Llama 2 and Stable Diffusion Read more…

Q&A with Nvidia’s Chief of DGX Systems on the DGX-GB200 Rack-scale System

March 27, 2024

Pictures of Nvidia's new flagship mega-server, the DGX GB200, on the GTC show floor got favorable reactions on social media for the sheer amount of computing po Read more…

NVLink: Faster Interconnects and Switches to Help Relieve Data Bottlenecks

March 25, 2024

Nvidia’s new Blackwell architecture may have stolen the show this week at the GPU Technology Conference in San Jose, California. But an emerging bottleneck at Read more…

Who is David Blackwell?

March 22, 2024

During GTC24, co-founder and president of NVIDIA Jensen Huang unveiled the Blackwell GPU. This GPU itself is heavily optimized for AI work, boasting 192GB of HB Read more…

Nvidia Looks to Accelerate GenAI Adoption with NIM

March 19, 2024

Today at the GPU Technology Conference, Nvidia launched a new offering aimed at helping customers quickly deploy their generative AI applications in a secure, s Read more…

The Generative AI Future Is Now, Nvidia’s Huang Says

March 19, 2024

We are in the early days of a transformative shift in how business gets done thanks to the advent of generative AI, according to Nvidia CEO and cofounder Jensen Read more…

Nvidia’s New Blackwell GPU Can Train AI Models with Trillions of Parameters

March 18, 2024

Nvidia's latest and fastest GPU, codenamed Blackwell, is here and will underpin the company's AI plans this year. The chip offers performance improvements from Read more…

Nvidia Showcases Quantum Cloud, Expanding Quantum Portfolio at GTC24

March 18, 2024

Nvidia’s barrage of quantum news at GTC24 this week includes new products, signature collaborations, and a new Nvidia Quantum Cloud for quantum developers. Wh Read more…

Alibaba Shuts Down its Quantum Computing Effort

November 30, 2023

In case you missed it, China’s e-commerce giant Alibaba has shut down its quantum computing research effort. It’s not entirely clear what drove the change. Read more…

Nvidia H100: Are 550,000 GPUs Enough for This Year?

August 17, 2023

The GPU Squeeze continues to place a premium on Nvidia H100 GPUs. In a recent Financial Times article, Nvidia reports that it expects to ship 550,000 of its lat Read more…

Shutterstock 1285747942

AMD’s Horsepower-packed MI300X GPU Beats Nvidia’s Upcoming H200

December 7, 2023

AMD and Nvidia are locked in an AI performance battle – much like the gaming GPU performance clash the companies have waged for decades. AMD has claimed it Read more…

DoD Takes a Long View of Quantum Computing

December 19, 2023

Given the large sums tied to expensive weapon systems – think $100-million-plus per F-35 fighter – it’s easy to forget the U.S. Department of Defense is a Read more…

Synopsys Eats Ansys: Does HPC Get Indigestion?

February 8, 2024

Recently, it was announced that Synopsys is buying HPC tool developer Ansys. Started in Pittsburgh, Pa., in 1970 as Swanson Analysis Systems, Inc. (SASI) by John Swanson (and eventually renamed), Ansys serves the CAE (Computer Aided Engineering)/multiphysics engineering simulation market. Read more…

Choosing the Right GPU for LLM Inference and Training

December 11, 2023

Accelerating the training and inference processes of deep learning models is crucial for unleashing their true potential and NVIDIA GPUs have emerged as a game- Read more…

Intel’s Server and PC Chip Development Will Blur After 2025

January 15, 2024

Intel's dealing with much more than chip rivals breathing down its neck; it is simultaneously integrating a bevy of new technologies such as chiplets, artificia Read more…

Baidu Exits Quantum, Closely Following Alibaba’s Earlier Move

January 5, 2024

Reuters reported this week that Baidu, China’s giant e-commerce and services provider, is exiting the quantum computing development arena. Reuters reported � Read more…

Leading Solution Providers

Contributors

Comparing NVIDIA A100 and NVIDIA L40S: Which GPU is Ideal for AI and Graphics-Intensive Workloads?

October 30, 2023

With long lead times for the NVIDIA H100 and A100 GPUs, many organizations are looking at the new NVIDIA L40S GPU, which it’s a new GPU optimized for AI and g Read more…

Shutterstock 1179408610

Google Addresses the Mysteries of Its Hypercomputer 

December 28, 2023

When Google launched its Hypercomputer earlier this month (December 2023), the first reaction was, "Say what?" It turns out that the Hypercomputer is Google's t Read more…

AMD MI3000A

How AMD May Get Across the CUDA Moat

October 5, 2023

When discussing GenAI, the term "GPU" almost always enters the conversation and the topic often moves toward performance and access. Interestingly, the word "GPU" is assumed to mean "Nvidia" products. (As an aside, the popular Nvidia hardware used in GenAI are not technically... Read more…

Shutterstock 1606064203

Meta’s Zuckerberg Puts Its AI Future in the Hands of 600,000 GPUs

January 25, 2024

In under two minutes, Meta's CEO, Mark Zuckerberg, laid out the company's AI plans, which included a plan to build an artificial intelligence system with the eq Read more…

Google Introduces ‘Hypercomputer’ to Its AI Infrastructure

December 11, 2023

Google ran out of monikers to describe its new AI system released on December 7. Supercomputer perhaps wasn't an apt description, so it settled on Hypercomputer Read more…

China Is All In on a RISC-V Future

January 8, 2024

The state of RISC-V in China was discussed in a recent report released by the Jamestown Foundation, a Washington, D.C.-based think tank. The report, entitled "E Read more…

Intel Won’t Have a Xeon Max Chip with New Emerald Rapids CPU

December 14, 2023

As expected, Intel officially announced its 5th generation Xeon server chips codenamed Emerald Rapids at an event in New York City, where the focus was really o Read more…

IBM Quantum Summit: Two New QPUs, Upgraded Qiskit, 10-year Roadmap and More

December 4, 2023

IBM kicks off its annual Quantum Summit today and will announce a broad range of advances including its much-anticipated 1121-qubit Condor QPU, a smaller 133-qu Read more…

  • arrow
  • Click Here for More Headlines
  • arrow
HPCwire