There is a major shift for enterprises migrating on-premise HPC to the cloud. 451 Research surveyed a panel of IT professionals to assess their intent on public and private cloud deployment. The results indicate an estimated 90 percent of companies will be in the cloud by 2020. There will be more buying and less building for organizations as 60 percent of IT deployments are to occur off-premise.
The chart below highlights the shift of execution environments today and two years from now. The projected trend shows a significant decrease of on-premise workloads and a migration to private and public cloud.
William Fellows, Founder and Research Vice-President of 451 Research discussed research findings on a webinar sponsored by Univa. He explains the industry is entering an era where cloud consumption overtakes cloud building. In this new deployment approach for IT resources, the infrastructure function becomes increasingly invisible to the end user.
“HPC and Cloud are two worlds which have been orbiting each other somewhat independently until recently,” said Fellows. “With the recent advances of both hardware and software, HPC is becoming much more mainstream. No longer is it only about 3D modeling and simulation, we are seeing a convergence of HPC across a variety of use cases such as smart cities, personalized medicine, fraud detection, autonomous vehicles, and augmented reality.”
The rise in the use of HPC companies taking advantage of containers is a key factor to the growing momentum. Applications and workloads are becoming blurred between Cloud and HPC. Traditional HPC apps are getting converted over for cloud-native and net-new cloud-native applications are being written specifically for HPC “big compute” environments. There are a wider set of choices now for businesses to process large data.
“Also, it is more affordable now for organizations to undertake the type of workloads that were unthinkable when they would have to go and purchase own CapEx IT equipment. Instead, they can just rent it,” said Fellows.
The cloud is also attractive to businesses where HPC is not a core skills focus, primarily because of the flexibility, instant availability, ease of upgradability, and security provided to run large scale jobs.
Fellows notes that cloud providers themselves are becoming HPC-enabled as demand increases for AI and machine learning resources, as well as support for analytics and big data. Specific workload requirements such as graphics, intelligent search and stream processing, has led to an explosion of development in new infrastructure powered by GPUs, FPGAs and specialist AI processors.
These dense computational architectures are becoming a requirement – resulting in the cloud and HPC moving in the same direction.
Click here to watch webcast on demand