Cloud computing as a business is fairly new. It was introduced by Amazon in 2006, followed by Google beta product in 2008, and last but not least – Microsoft released its first version of Azure in 2010. Since then clouds have been evolving at an astonishing pace, constantly improving performance, inter-communication, and adding new services. One of the very recent additions to cloud services is HPC (high-performance computing). Due to the advanced networking capabilities introduced to the data-centers, combined with ever more powerful servers, clouds are capable of creating ad-hoc clusters with a massive number of servers, essentially providing supercomputer services on demand. One of the main advantages of such HPC clusters for the user is the pricing model. Users pay only for what they use. There is no fixed cost of constructing or renting a specialized data-center building. Moreover, there is no need for either buying or installing the expensive hardware and keeping an IT department on the payroll. Most importantly: once the cluster has finished its job, it disappears without incurring further expenses to the user.
While it is true that dedicated clusters built the “old-fashioned” way provide a better performance, the cost of the cloud cluster makes it an undeniably attractive option.
There are three major cloud services providers today: AWS (By Amazon), Azure (by Microsoft) and Google. AWS is the current leader, with Azure breathing down its neck, and Google somewhat lacking behind, but still a formidable competitor. There are other providers like IBM Bluemix, Alibaba Cloud, and Digital Ocean, to name just a few. However, the vast majority of the cloud business is held in the capable arms of the aforementioned tech giants.
It is no trivial task to develop a product on one cloud platform. However, it is a good strategy not to be bound to any one service provider. This is true for any business and applies to cloud providers as well. Due to the competition between the cloud providers, one can never know with certainty which provider will come up with a new revolutionary service or a more attractive pricing model. Therefore in order to keep the cost of HPC low, it is recommended that businesses use more than one cloud provider (i.e. AWS, Azure, and Google) assuming they all provide the required capabilities for the business’s needs.
Developing product on the cloud platform(s)
Developing a product on the cloud platform requires unique expertise and skills. It takes a dedicated team of engineers specializing in multiple cloud platform aspects to successfully use cloud HPC. Clouds evolve on a constant basis – which means one must continuously track the changes and assess their impact on the cost and performance of using an HPC cluster. New types of servers are introduced to data centers resulting in new services and new pricing models. This continuous evolution implies that Cloud development is an ongoing task which requires a fully dedicated R&D team.
At Sphere, our mission is to make the cloud HPC accessible to any company, without having to keep its own cloud R&D team on the payroll. We completely offload all cloud HPC development and maintenance. Sphere provides HPC performance with (multi)cloud flexibility, at an optimized cost and minimal effort on the part of the user, which allows our clients to concentrate on their core business.
High-Performance Computing Use Cases
High-Performance Computing Use Cases – BI services
BI software may require running complex computing and data-intensive models. Clients of the BI service may choose to run the software on-premise using the local PC(s), however, the time required for analysis is often crucial and will take too long unless running on a big cluster of servers. In that case, if the BI software is integrated with HPC, the client may choose using the cloud service. If not already in place, the relevant data is uploaded to the cloud and the model is run on the required number of servers in parallel. From the client’s perspective, they are using BI software. but “under the hood” – BI software is using the cloud. Financial services which provide market analysis for example, could be adapted to HPC in a similar manner.
High-Performance Computing Use Cases – FinTech
Fintech companies have historically been the early adopters of advanced computing technologies. The fintech industry requires powerful computing power in order to analyze complex financial models, with short turnaround times. Sphere Consulting services are a dream come true for Fintech companies that are seeking to speed-up their development cycle, while keeping R&D and infrastructure costs at bay.
High-Performance Computing Use Cases – Physical modeling & 3D printing
3D printing requires running complex simulations of the physical world which in turn requires extensive time, even using the most powerful traditional servers. By connecting the local software, the user of the 3D modelling software may choose to run the model on the cloud. Going with this option will seamlessly connect the software running on the user’s local server and run the model on a cloud HPC cluster of the required size.
High-Performance Computing Use Cases – Computer graphics
Computer graphics and especially 3D graphics are very compute-intensive tasks, making them ideal candidates for parallelizing and processing on an HPC cluster. Offline rendering can be split into parallel tasks and run simultaneously on an HPC cluster. Similarly to the previous use case, rendering could be seamlessly offloaded to the Sphere Consulting platform, thus reducing the time required to complete this task.
High-Performance Computing Use Cases – Research / Pharma
There is an infinite number of use cases for HPC in research and specifically in the pharma domain. For example, the BAC (Binding Affinity Calculation) problem requires enormous computing power and is a natural fit for HPC. Sphere Consulting allows Pharma companies to cut costs by using cloud HPC and to concentrate on research rather than spending valuable resources and time on development.
Sphere Consulting has a relentless focus HPC and the client’s experience around cloud HPC. We remove the high expertise threshold required to enjoy the HPC benefits and make it accessible to our clients without the distraction of infrastructure maintenance, or complexity of the parallel programming paradigms.