Not many technological trends are currently receiving more hype than artificial intelligence and machine learning. This should come as no surprise – nowadayseven the smallest devices, whether smart watches or industrial sensors, produce increasing amounts of data, and advanced data science methods quickly find application in the analysis of growing data sets. The features of IBM POWER9 processors make them top performers when it comes to advanced analytics. With the advent of POWER-based cloud platforms, advanced AI and ML analysis can become even faster.
There is more to data than this
In 2017, The Economist published a report titled “The world's most valuable resource is no longer oil, but data”. Behind this catchy phrase lies both truth about current economic trends, and a certain simplification about value of data. Indeed, less than 10 years ago Exxon Mobil and PetroChina were dethroned as business leaders, and currently five of the best-earning companies are IT giants – the FAAMG group - famous for their data collecting and game changing data-driven marketing strategy. At the same time, it seems like we are still in the pre-Rockefeller era of data business. Advanced data analysis methods such as machine learning and deep learning provide great results for advertisement re-targeting and positioning best-suited content on social media, and work extremely well in financial services. Yet we are yet to see the revolutions of the future, which can be delivered by today’s technology. Gartner states that the digital business model era will be followed by the algorithmic business model era, focused on mathematical operations and providing a competitive advantage by solving complex challenges of manufacturing, medical diagnosis and predictive analysis – and we the right tools for these tasks.
POWER – designed for data
Such challenges require special features from the hardware that hosts them. This is where IBM Power Systems – a series of servers utilizing processors built on POWER architecture as an alternative to traditional x86 architecture that dominates the industry – come into play. Power Systems technology originated decades ago, and was created from the start to treat large amounts of data efficiently. This core mission remains true today, in IBM’s POWER9. Power Systems are designed with focus on data-driven workloads handled by enterprise-class servers, as opposed to supporting multiple devices by the multiple design versions of the x86 family. The biggest advantages of POWER architecture’s design are its acceleration capabilities and compatibility with advanced technology, such as POWER9’s industry-exclusive CPU to GPU high speed NVIDIA NVLink interface, PCIe Gen4 for non-volatile storage access, and the OpenCAPI interface, which offers high bandwidth communication to NICs and FPGA accelerators. The technology allows impressive results when benchmarking the performance of ML and AI algorithms against their x86 counterparts, achieving four times the number of threads per core and up to 9.5 times more memory.
Power Systems are currently used whenever superior performance and availability are required – whether in banking, insurance or logistics. Of all Fortune 500 companies, 80% run their most critical systems on the platform. POWER9 quality features bring ever more big data and data science applications to the platform. During last OpenPOWER Summit, Google confirmed that it has deployed the "Zaius" platform at its data centers for production workloads. Industry giants such as Alibaba and PayPal use POWER computational force to perform transactional and large dataset processing tasks, with Tencent stating that it saved 30% of resources and increased efficiency by 30% by introducing Power Systems at its data center.
A change is in the air
But why is it all so important right now? The year 2019 is more exciting than ever for Power Systems, which is down to one word – cloud. Up to this year, Power Systems were available only as on-premises products managed by in-house IT teams or outsourced to a partner company on a managed services basis, usually as a Capex investment. This approach was not always efficient, as decisions about the appropriate system size requirements are rarely pinpoint when made in advance. Additionally, data scientists have been limited by infrastructure issues, lacking hardware resources when running cognitive workloads such as AI or deep learning. In June 2019, IBM announced the launch of its cloud platform based on Power Systems, but it is worth mentioning that Comarch, the company’s technological partner, already has its own PowerCloud solution up and running.
Comarch PowerCloud is a comprehensive solution allowing users to move their infrastructure into cloud in a way most convenient for them – whether they need private, public or hybrid cloud products. Thus, any company can bring its business up to speed without having to worry about security and system management – for Comarch provides experienced engineers to maintain hardware and software. Equally important is the pay as you go model, which is in demand as a means of cutting and controlling costs. With this Opex investment strategy, companies get access to top-notch servers and mass storage for a fraction of the cost of in-house systems – with no entry or exit barriers. In the experience of Comarch’s specialists, transferring the system to the cloud may reduce costs by as much as 30-40%. Easy scaling of the cloud-based systems helps overcome the “infrastructure wall”, which traditionally imposes limitations that prevent maximum speed algorithmic work. This slowdown isn’t good news, because high-speed operations are a vital part of any big data project. With Comarch PowerCloud platform, companies are able to run PowerAI or Watson Studio applications without worrying about maintaining their Power Server or being held up by limitations of on-premises infrastructure. Considering the performance quality of POWER processors, we can expect nothing less than the extraordinary from applications run on PowerCloud.
To find out more about transferring your Power System into the cloud, and to keep updated with current development, feel free to follow
and check out
Product Manager Comarch ICT