Cloud computing takes shape
HPC users are increasingly turning to cloud technologies due to their flexibility and scalability, allowing them to quickly change the size of their workloads, adopt new technologies in a small testing environment and to help to increase the agility of a company working across multiple sites, or even countries.
However, there are many different options from managed services to on-premise cloud installations and hybrid approaches, which each offer their own benefits to different user groups.
A recent survey carried out by Ansys and Digital Engineering aimed to identify best practices and insight into the type of cloud technologies being adopted by HPC and simulation users.
The survey found that the cloud is beginning to take hold of the simulation user landscape. For example, 17 per cent of respondents are currently using the cloud and another 20 per cent plan to do so within 12 months.
‘The growing need for cloud options is fuelled by the trend of users increasing the size of their simulations. These large simulations require more high-performance computing and data storage resources,’ states Wim Slagter, director for HPC and Cloud Alliances at Ansys.
The survey notes that more than a third of users reach peak computing capacity a few times per month. When these moments occur, users can access cloud resources to meet the need for additional storage and HPC.
Respondents were requested to note if they worked for a large, medium or small company. Based on those responses, 42 per cent of those that are already using the cloud for simulation purposes are working for larger companies – while those working for medium and small companies split the remainder at 29 per cent each.
However, smaller companies appear to be adopting the cloud faster than others, with 44 per cent of respondents working for small companies reporting that they plan to implement the cloud in the next 12 months – compared to those working for large- and medium-sized companies report 35 per cent and 21 per cent respectively.
‘If an engineering business is going to keep up with the trends, it needs to start adopting simulation-enabled cloud services sooner rather than later. In general, respondents are saying the companies they work for are engaged in cloud-based solutions, because they are looking for flexibility and lower cost,’ added Slagter.
Survey respondents reported the three main reasons why companies are moving to the cloud are; the ability to scale up quickly, lower overall costs and lower maintenance costs.
‘Considering the reasons companies are adopting the cloud, any firm lagging behind competition should start to implement some fundamental cloud best-practices and adoption procedures,’ said Slagter.
Firms that have not yet chosen to adopt cloud-based simulation tools can look at some of the best practices that have been discovered by the study. The top three cloud best-practices noted by the survey’s respondents were; test and share potential time savings with decision-makers, address concerns such as security early in the process and the ability to start with small cloud investments.
‘Engineers will need to perform many tests to make sure that decision-makers clearly see the benefits of the cloud.
‘They will need to understand how the cloud saves time and how that translates into productivity gains, increased throughput, reduced time-to-market and cost savings.
‘Many respondents noted that having lofty goals during the start of a cloud simulation initiative can undermine the entire enterprise. They suggest finding small simulation projects that have been difficult to support with internal resources. Then, launch a pilot to prove that the cloud has value. Engineers who target the pain points of these particular projects should quickly gain support to expand the cloud to other systems,’ added Slagter.
Moving to the cloud is not a one-size-fits-all solution to computing and engineering challenges. This is why the survey and the wider cloud market, in general, find users are adopting a mix of on-premise, partner-managed and hybrid cloud approaches.
Slagter predicts that over the next 12 months that mix will continue to change, as more users begin to adopt managed solutions. This will be in the form of either managed software-as-a-service (SaaS) from an independent software vendor (ISV), where the software provider also provides the cloud infrastructure; a combination of a private cloud and an ISV-managed SaaS solution; a partner-managed cloud using public data centres or a combination of private cloud and partner-managed cloud.
‘Many companies are already capitalising on cloud-enabled simulation and a growing number of others are preparing to do so shortly. With on-demand access to HPC resources, these organisations have been and will be, able to accelerate their simulation activities,’ concluded Slagter.
Featured case study: Combatting epidemics with synthetic biology and cloud computing
Infectious diseases continue to plague populations worldwide. While vaccination has proven to be exceptionally powerful to counter this threat and some infectious diseases have been eradicated or constrained, severe challenges persist with epidemics caused by Ebola, Zika and others, particularly in developing countries.
A recent example is Chikungunya, a virus transmitted by the bite of an infected mosquito. The disease causes crippling headache, vomiting, swelling of limbs and can lead to death.
Formerly confined to sub-Saharan Africa, Chikungunya has recently spread, as its mosquito host leaves its natural habitat due to deforestation and climate change, with recent outbreaks in the USA and Europe.
Researchers from the University of Bristol and the French National Centre for Scientific Research, teamed up with computer technology giant Oracle and vaccine innovator startup Imophoron to find a way to make vaccines that are thermostable (withstand warm temperatures), can be designed quickly and are easily produced.
The research resulted in a new type of vaccine that can be stored at warmer temperatures, removing the need for refrigeration, in a major advance in vaccine technology.
The findings, published in Science Advances, reveal exceptionally promising results for the Chikungunya vaccine candidate, which has been engineered using a synthetic protein scaffold that could revolutionise the way that vaccines are made and deployed.
To validate their design, scientists employed cryo-electron microscopy, a powerful new technique, in Bristol’s state-of-the-art microscopy facility headed by Christiane Schaffitzel, co-author of the study. Cryo-EM yields very large datasets, from which the structure of a sample can be determined at near-atomic resolution, requiring massive parallel computing.
Enabled by Oracle’s high-performance cloud infrastructure, the team developed a novel computational approach to create an accurate digital model of the synthetic vaccine.
University of Bristol IT specialists Christopher Woods and Matt Williams, together with Oracle, implemented software packages seamlessly on the cloud, in this pioneering effort.
Woods explained: ‘We were able to process the large datasets obtained by the microscope on the cloud in a fraction of the time, and at a much lower cost, than previously thought possible.’
Phil Bates, leading cloud architect at Oracle, said: ‘Researchers have had a long tradition of building and installing their own supercomputers on-premises, but cloud computing is allowing them to run large datasets in record time, with fast connectivity and low latency.
‘This helps them crunch data and make scientific breakthroughs faster. Technologies like machine learning and cloud computing will play a significant part in the scientific world, and we are delighted we could help the researchers with this important discovery.’
‘We were thoroughly delighted,’ said Imre Berger, director of the Max Planck-Bristol Centre for Minimal Biology in Bristol. ‘Our vaccine candidate is easy to manufacture, extremely stable and elicits a powerful immune response.
‘It can be stored and transported without refrigeration to where it is most needed. Intriguingly, we can now rapidly engineer similar vaccines to combat many other infectious diseases just as well.’
Fred Garzoni, founder of Imophoron, said: ‘Matching cutting-edge synthetic biology with cloud computing turned out to be a winner.’
High Performance Computing (HPC) on Oracle Cloud Infrastructure is enabling scientists and researchers to do more science. Bare Metal compute enables on-premises performance in a flexible cloud environment for HPC applications. Compute, networking and storage are all based on this high-performance paradigm and provide performance at prices that are competitive with on-premises hardware.
HPC pushes the boundaries of compute, networking and storage. Oracle offers compute built on proven Oracle Servers and processors from Intel, AMD, and Nvidia. If the workload requires thousands of cores or specialised processors, Oracle Cloud can accelerate that workload.
Many tightly-coupled HPC workloads rely on ultra-low latency networks. Oracle provides a stable RDMA network that delivers on-premises latency across thousands of cores. In addition to RDMA, the network for non-RDMA traffic is flat and non-oversubscribed.
For bare-metal HPC computing this means no noisy neighbours and predictable performance between nodes, racks and data centres.
Oracle Cloud object and block storage is built on the same high performance compute and networking, which means that there is no additional charge for performance, and high performance file servers can be built on the cloud with performance similar to on-premises.
Research, simulation, and engineering high performance computing workloads all over the world are migrating to the cloud seamlessly. Oracle Cloud provides high performance and more flexible costs that can be directly matched to your specific workload.