ERP managers are finding more reasons to turn to high-performance computing to improve the performance of big data analysis and machine learning models.
ERP systems have an enormous amount of data that many firms are just beginning to tap. To answer complex questions with this information -- and to do so quickly -- it takes a lot of computing power. It is one reason why Oracle just added a high-performance computing, or HPC, cloud capability to its portfolio of public cloud infrastructure services.
The new Oracle HPC cloud capability is aimed at two audiences. The first is users who run legacy HPC applications on premises, such as scientific projects, R&D applications and virtual product design in lieu of physical prototypes. The second audience is ERP managers who need to apply high-end computing resources to business data processed in the Oracle Cloud Infrastructure service.
ERP systems manage a lot of data. "Most of our customers today are trying to find more and more value out of that data, and that's what we're we are seeing [as] growth," said Karan Batta, director of product management for Oracle Cloud Infrastructure.
HPC expanding to mainstream business uses
The use of HPC outside of its traditional scientific and research use is expanding to areas such as sales analysis and planning, supply chain planning and HR-related workforce analysis, said Steve Conway, senior research vice president at HPC research firm Hyperion Research in St. Paul, Minn.
An HPC system can take in far more data to allow for deeper analysis. This approach is also being coupled with AI technologies, such as an inference engine that can be applied to many situations in ERP, Conway said.
"The larger the data sets, the more accurate the results are going to be," he said.
Enterprises not only want ERP-related questions answered "that are more complicated than before, but they also want the answers in something very near-real time," Conway said.
HPC ideal for machine learning models
The larger the data sets, the more accurate the results are going to be.
Steve Conwaysenior research vice president, Hyperion Research
High-performance computing is a set of technologies and processes designed to maximize performance. What Oracle HPC offers users is a clustered network with access to bare-metal processing, both CPU and GPU. Bare metal is a single-tenant server or system that doesn't use virtualization, which can add latency.
Another key technology Oracle HPC uses to speed performance is remote direct memory access (RDMA), which allows an application to write directly to memory remotely without involving the CPU or operating system.
The use of bare metal and RDMA in a cloud platform means vendors are "getting past one of the very big bottlenecks that affected a lot of cloud computing, which is virtualization," Conway said.
Business application developers can take mammoth data sets from ERP systems and put this data into a machine learning model. From that model, a business can figure out what "kind of insights they can generate out of that data," and then feed it back into the ERP system, Batta said.
Machine learning models are computationally intensive and can take hours, days or even weeks to run if they don't have access to enough computing resources, Batta said. That's where high-performance computing comes in.
The Oracle HPC system will allow scaling up to 1,080 cores for a single project, although the number of available cores will be expanded in time, Batta said.
