After a number of very strong years, 2014 ended soft. The Top5 system purchases are slowing down for about two years now. The IBM/Lenovo deal has delayed many purchases but Earl Joseph stated that the upcoming years 2015 to 2018 are predicted to be healthy growth years.
Big Data combined with HPC is creating new solutions but the software issues will continue to grow. GPUs and accelerators are also extending their impact. The use of Clouds is growing. The sites currently using Clouds amount to 31,2%.
The IDC analysts thought that non-x86 processors could alter the landscape. China looms large and there is a growing influence of the data centre in the IT food chain. HPC in the Cloud is equally gaining traction.
Storage is fastest-growing part of the HPC market with revenues growing to record levels but the storage market remains fragmented. The HPC interconnect market is in transition. The movement and mangement of data become a major pain point. On the other hand, HPDA will boost the storage budgets.
As far as the competitive segments in 2014 are concerned, except for the top10 systems, the market grew dramatically. The big news is that the IBM/Lenovo split-up has changed the market, with IBM now representing 4,7% and Lenovo 15% of the market. HP is the number 1 with 36,1%, representing one third of the market. The IBM decline is higher than expected. The supercomputer segment underwent tremendous shifts in 2014. The financing sector grew by 50%. The IDC analysts forecast a yearly growth of 8,2% in HPC from 2014 to 2019. In 2019, $15,2 billion should be reached.
In the European perspective, the HPC market has declined in the past but is now experiencing a tremendous recovery.
IDC is completing a new study for the European Commission that is seeking to invest in an exascale machine. There will be an increase in the amount of supercomputers in Europe. In 2020, a purchase of 2 exascale machines has been planned and in 2022, a purchase will follow of another 2 exascale machines.
IDC has a large research programme around High Performance Data Analysis (HPDA). In this programme there are four new categories: fraud and anomaly detection, business intelligence, marketing, and other commercial HPDA.
Typical machine learning is working on bigger data sets. For HPDA advanced simulation and newer HP analytics are required. IDC predicts a fast growth from a small starting point. HPC and high-end commercial analytics are converging and the algorithmic complexity is the common denominator. Economically important use cases are already emerging. In any case, there is no single HPC solution available that is best for all problems.
In HPC ROI research, IDC received a grant from the Department of Energy last year to provide a 3 year research plan. In 2013, a pilot study was issued in which three approaches have been tested and models have been set. The models have to be refined to move from association towards causation. The ROI models that have been developed include an ROI based on revenues generated divided by HPC investment; an ROI based on profits generated divided by HPC investment; and an ROI based on jobs created.
IDC has also developed two new index scales to measure scientific investigation and discovery. The first scale measures the quality of innovation and the second scale measures the impact of the innovation.
IDC has a mission to improve the health of the HPC industry by organizing HPC User Forum meetings, together with an award programme. IDC has organized over 50 meetings worldwide since 2000. More information about the HPC User Forum meetings is available at www.hpcuserforum.com .
In July 2015, there have not been granted any IDC HPC innovation awards. These awards serve to help expand the use of HPC by showing real ROI examples. Some of the submissions described work that seemed both important and interesting but none of them included information that was sufficient enough to permit fair judging. Candidates can apply again for the fall 2015 award. IDC is currently updating the application and review process for the HPC innovation awards to increase the quality of the submissions. The new guidance for future submissions will be ready, starting with SC15 in November.
Earl Joseph and Bob Sorensen concluded their talk by explaining why HPC is projected to grow. The low half of the market is finally back to a recovery mode. HPC has become a competitive weapon. Governments view HPC leadership as critical for economic prosperity. However, there are very critical HPC issues that need to solved, including global warming, alternative energy, safe NE, financial disaster modelling, health care, homeland security, etc.
In HPC, there are still major customer pain points. Software is the no. 1 roadblock. Parallel software is lacking for most users. Clusters are still hard to use and manage. Power, cooling and floor space are the major issues. ROI is becoming a requirement, especially as system costs escalate. Fortunately, there are new technologies in the Big Data, accelerator, and Cloud field.
In any case, HPC is still expected to be a strong growth market but the vendor share positions will shift greatly in 2015. The recognition of HPC's strategic and economic value will drive the exascale race. The formative HPDA market will expand opportunities for vendors. The high ROI will drive major segments, such as health care and finance. China will definitely be a wild card with the US-China supercomputer battles that continue to go on.