Three months ago, SGI announced the development of DataRaptor, its big data system powered by the MarkLogic database. The system was introduced and explained by Steve Conway, HPC Research Vice President of IDC, and Bill Mannel, SGI’s VP of Product Marketing, in a talk recently summarized here.
Mannel and Conway talked again this week about SGI and their various big data functions. Their focus here was specifically data-intensive high performance computing, otherwise known as high performance data analytics (HPDA). The talk touched on both the rise of big data within HPC and how SGI has molded to that.
According to Conway, HPDA is less about asking a general question to find a specific answer, or finding a ‘needle in a haystack,’ and more about generating new questions that have not yet been asked whose answers may not exist in the metaphorical haystack.
This difference can be well-illustrated in the healthcare industry. Professionals may one day input patient symptoms into a system powered by an expansive database to help determine the proper procedure. That answer likely already exists in the database such that the system only has to relay that information instead of generating new data.
On the other hand, according to Conway, the government runs five major healthcare programs and loses billions of dollars in fraud from those institutions each year. An HPDA system must find the correlations among the data in its vast databases and generate new queries to investigate and capture creative fraud methods.
Fraud detection is mentioned in almost every lengthy discussion hosted about big data and as such may be somewhat of a tried subject. However, a system that could recover even a quarter of the funds lost by healthcare would be lucrative, as current studies estimate that less than a billion per year is actually caught.
Therefore, as noted by the table below, growth in data-intensive high performance computing and its place in the general HPC market has steadily grown over the last few years and is projected to continue to do so.
According to Mannel, SGI fits nicely with that growth, as their portion of revenue from big data hovers between 8 and 10 percent, a higher percentage than IBM, Dell, or HP. However, statistics like this without context can be a tad misleading, as only a percentage of revenue and not total revenue is discussed. The intent here is to show SGI’s increasing focus on the big data marketplace. An outline of SGI’s view of the big data world is illustrated below.
The figure above is a busy one, encompassing several data sources from social media to reconnaissance and research. As shown, their “UV Big Brain” gathers fast data from those sources and transmits it to ‘end nodes,’ where the DataRaptor system would help users make those free-form connections mentioned earlier.
The opportunities in high-performance big data are there, especially with regard to lingering fraud detection problems. Whether SGI’s vision for conquering those opportunities proves fruitful will be something to watch over the coming months and years.