The representatives of the cloud platform FlyElephant in collaboration with Squadex displayed their abilities in high-performance computing and machine learning at the ISC High Performance 2018 event that took place in Frankfurt, Germany on June 24-28.

Dmitry Spodarets (CEO at FlyElephant) and Aleksandra Spodarets (Business Development at FlyElephant) were at the AWS booth because FlyElephant is now Advanced Technology Partner, showed the demo of the FlyElephant platform and gave valuable advice about ML and HPC to the attendees.

At ISC 2018, we demonstrated how data scientists could quickly create working environments using their required software and the powerful Amazon EC2 C5 or Amazon EC2 P3 instances for developing or automatic training of machine learning models. We exemplified that using a model for image classification.

Using the special case for educational institutions, we shared some details on how the FlyElephant platform was deployed on AWS for simultaneous workloads of more than 400 students who did ML laboratory work.

On top of that, we presented to attendees and prospects how SQUADEX augments the FlyElephant platform with their services. For instance, FlyElephant would be supported by the SQUADEX consulting team to enhance the development and optimization of machine learning models. We are looking forward to improving our service on the Data Science tasks to help our clients. By joining forces, our target is to help businesses get the most out of Data Science with the help of machine learning algorithms and HPC technologies.
Dmitry Spodarets, CEO at FlyElephant

FlyElephant teamed up with SQUADEX before the ISC 2018 event. The goal of the collaboration is to empower FlyElephant platform with top-notch consulting services and introduce new AI- and HPC-enabled solutions to the market.

But, without further ado, let’s get back to the conference.

What is ISC 2018 event about

ISC High Performance is the HPC event that brings together experts, engineers, and ecosystem-minded leaders with the variety of specializations – from HPC and AI to climate change and cosmology.

This year, ISC High-Performance event gathered over 3,500 visitors who came from over 60 countries (70% of attendees and exhibitors are not from Germany). 162 companies and research institutions had a chance to announce and present the latest technologies and services in the industry.

To keep focus on high-performance computing in scientific and commercial fields, ISC 2018 kept the balance between science and industry participants of the event. They organized specialized sessions to reveal specific aspects of HPC. Artificial intelligence, machine learning, and industrial technology sessions were conducted as well. The event follows the important movements and trends, such as global warming, climate change, gender equality, and even the FIFA World Cup.

Announcements and Rising Trends

The Update of Top500 Supercomputer Rankings

The International Supercomputing Conference 2018 kept an eye on the update of the Top500 Supercomputer rankings. The IBM-built supercomputer “Summit” (USA) took the top spot.

The top 5 is under big rearrangement. Four supercomputers are either new or got a huge update. The top 5 for June 2018:

Source: top500.org

Summit consists of 4,356 nodes, it is powered by two 22-core Power9 processors and six NVIDIA Tesla V100 GPUs. Summit is able to deliver 122.3 petaflops per second.

The other systems included in the top 5 are clearly demonstrating the usage of CPUs and GPUs for augmentation instead of other hardware.

Source: Mellanox June 2018 Top500 analysis slides

HPC Moves to the Cloud

Addison Snell, CEO of Intersect360 Research, reported that HPC in the cloud grew by 44%, comparing to 2017 in the dedicated industry report.

Source: Intersect360 Research

Estimating to reach 1.1 billion US dollars by mid-2018, the sector is forecasted to get the biggest boost in 2019 and eventually will reach 3 billion US dollars by 2022.

Snell refers to the growth of HPC cloud computing to two factors:

  • The widespread use of ML in the cloud. Machine learning apps are set up to be used in the cloud by default.
  • Simplification of license migration to the cloud. Licensing models are becoming more advanced and provide easy to use options to developers.

The total amount of cloud HPC jobs will be about 15-20% by 2020.

AI & ML Development Within the HPC Domain

AI & ML are about HPC, though there were just a couple cases of their successful usage with improving performance, scalability, and configurability of real-world solutions before 2018.

AI, ML, and Big Data (HDPA) are likely to be used along HPC to improve a product’s architecture which was demonstrated by various apps at the ISC 2018. It is optimizing the use of computer resources, and to significantly boost a system’s overall performance.

This combination may cause the development of broad capability systems with next-level configurability and cross-pollination. Such kind of systems will have requirements for a new set of performance metrics, completely different from Linpack.

HPC Space Keeps Growing

According to Earl Joseph, CEO of Hyperion Research, high-performance computing grew by 10% to reach 24.3 billion US dollars in 2017.
The sector is estimated to be constantly growing till 2022, by 10% year-on-year, and is set to reach 38.4 billion US dollars.

The growth will be directed by AI & ML, high-performance data analytics and HPC in the cloud. The government will also participate in the next-gen exascale supercomputer systems financing.

Addison Snell’s forecast for HPC Product & Services (up to 2022):

Conclusion

FlyElephant is proud to be the part of the International Supercomputing Conference High Performance 2018. We introduced FlyElephant and SQUADEX  as really efficient service providers in ML, HPC, and Big Data spaces.

We paid attention to significant and important trends that will definitely impact the HPC landscape in the future.

Artificial intelligence, machine learning, and HDPA will tie and strengthen up each other. The enhancement of AI, ML, and Big Data in the HPC cloud space might cause a major update of Linpack metrics, focusing on more flexible, real-time indicators of performance. HPC in the cloud will keep growing, both service- and job-wise; HPC spending and revenues will gain momentum, with a steady growth of 10% yearly up to 2022.

FlyElephant will keep you updated about upcoming conferences and events.

Stay tuned!