Earl Joseph, Vice President of HPC at IDC

Supercomputing 2015 was the opportunity to hear from the server vendors, that the business sector has the highest market growth. This trend has contributed to the growth of the total HPC server market by nearly 12% to $ 11.4 billion at the end of 2015, according to IDC. Which is an upward revision of the forecasts IDC had at the beginning of the year in part because of the rapid adoption of server computing in the financial services sector. “The combination of HPC and Big Data creates new solutions,” said Earl Joseph, Vice President of HPC at IDC, “with particularly strong growth in the simulation and analysis of data-intensive, and adding many new users and purchasers in the HPC industry. ” Among the figures, those who most caught our attention concerned the ROI, return on investment. According to the latest data from IDC, every dollar invested in the HPC sector generated an average 514.7 dollars in revenue. IDC estimates that this growth will be sustained next year if not amplified by the problems related to storage, strong growth also. The overall turnover of the HPC global level should be 22.1 billion. The combination of Big Data and HPC continues to induce changes in the way IDC defines and monitors this emerging market. To qualify, IDC coined the term High Performance Data Analysis (HPDA). Joseph Earl notes that the HPDA brings out new solutions and added many new users and buyers in the HPC market. The financial sector, for example, “has grown faster than what we measured in the last two years (about 50% higher). ”

Currently, IDC has identified four sub-segments for HPDA:

Fraud detection – This segment focuses on the identification of suspicious or potentially suspicious behavior and their causes using the graphical analysis, semantic analysis, or other high-performance analysis techniques.

Marketing – This segment covers the use of HPDA to promote products or services, generally using complex algorithms to discern the demographic characteristics of potential customers, their preferences and consumption habits.

Business Intelligence – This segment uses HPDA business to identify opportunities to advance their businesses and business competitiveness through a better understanding of their business, their competitors and the changing dynamics of the markets in which they operate.

Commercial HPDA – This segment includes all commercial HPDA workloads other than the three described above.

Over time, IDC expects some of these workloads can become important enough to become full-fledged sub-segments. For example, the use of HPDA to manage large IT infrastructures, and the Internet of Things. A future segment HPDA said Joseph, will be precision medicine. A sample application will focus on the analysis of medical outcomes for diagnostic purposes and treatment planning. In this paradigm, history and the patient’s symptoms are recorded in a database. While the patient is still in the office of his doctor, the latter will be able to correlate the millions of archived records of patients with similar symptoms to get relevant results. The doctor will be able to take into account the rate of effectiveness of different treatments for patients with similar disease profile, without being bound by the conclusions. In this, the HPDA applied to medicine can be a powerful decision-making tool. Public and private payers have long encouraged the development of similar approaches to evidence-based medicine in which entire populations of patients records could be examined and evaluated to determine which drugs and therapies are useful and should be approved. In theory, the result in both cases would be a better success rate and a reduction in aberrant and costly practices. The cloud-based HPC computing is on the rise in both volume (we went from 13.8% of sites in 2011 to 23.5% in 2013 and 34.1% in 2015), as is the number of running workloads.

Other findings of the IDC report are:

• The growing recognition of the strategic value of HPC helps boost premium sales. The entry-level buyers are also back in a growth mode.
• Marketshare positions of HPC providers will change significantly in 2015.
• Recognition of the strategic /economic value of HPC stimulate the race to exascale, with systems capable of 100 PetaFLOPS power becoming commonplace in 2016. exascale infrastructure should arrive between 2022 and 2024.
• Non-x86 processors could change the landscape – Power processors, but also ARM coprocessors, GPUs and FPGAs.
• China will occupy a more important place. Lenovo, benefiting from the growth of the domestic market, will confirm its export intentions. Other Chinese suppliers are planning to expand to Europe.
• The growing influence of the data center in the IT channel will impact technology options open to the HPC ecosystem, perhaps by providing new approaches.
• The HPC in the Cloud is gaining ground and the big questions are how far and how fast

.

Of course there is still room for improvement. The software remains a major obstacle. Improved management software and HPC resource administration is essential and there is a persistent lack of parallelized software for the majority of users. In addition, many applications would need an overhaul or a major rewrite in order to function in HPC environments. Clusters are still notoriously difficult to deploy, use and manage. The issues of power, cooling and floor space remain major challenges. Moreover, IDC indicates that HPC still suffers from a lack of support in heterogeneous environments combining processors and accelerators. Finally, storage and data management are growing bottlenecks given the foreseeable significant increase in data volume.