Many HPC aficionados probably think of Enterprise Computing as something static and boring: a solved problem; something to be maintained and occasionally updated; or maybe moved to a Cloud – but not a fruitful area for novel approaches or exotic hardware. Big Data may change those views. Let’s take a look.
Enterprise Computing
What is Enterprise Computing? Wiktionary defines enterprise as “A company, business, organization, or other purposeful endeavor.” In their enterprise computing text, Shan & Earle state that “Enterprise computing involves the development, deployment and maintenance of the information systems required for survival and success in today’s business climate.”
OK, so far so good. It would seem that enterprise computing is a function of the nature of the purposeful endeavor and the business climate in which it is immersed.
Business Climate Change
The emergence of Big Data is clearly changing the general business climate. This point was made clearly and compellingly at the recent Big Data’13 conference. The traditional structured data used in business intelligence, stored in the rows and columns of countless spreadsheets and in SQL databases, is being rapidly augmented by large and rapidly growing volumes of unstructured data. Google alone is said to be acquiring unstructured data at the rate of a petabyte per hour. The expectation that competitive advantage can be had by more effectively and more rapidly using all of this structured and unstructured data is now widely held. This has driven business intelligence into big data analytics – including graph analytics – and more aggressive, complicated and resource-intensive use cases. These new aspects of business intelligence, in turn, create a demand for new applications, algorithms and computing system architectures. Starting to sound like our world of HPC?
Big Data is HPC
We have previously observed that Big Data is a form of HPC and should be embraced as such. This is currently happening in several science disciplines, such as high energy physics, astronomy, and biology. The need for data analytics and, in particular, visual analytics is driving Big Data-as-HPC into additional sciences – including human health. The emerging Internet of Things will make Big Data much bigger, more valuable and useful in new ways – further embedding Big Data and data-intensive computing into HPC.
A recent presentation to the Secretary of Energy’s Advisory Board clearly indicates that the advocates of the Department of Energy’s Exascale Initiative understand that the future of high-end computing lies both in Big Compute and Big Data. So, it’s reasonable to expect that future applications, algorithms and computing architectures for science will be developed to serve both aspects of HPC.
At this point in the development of Big Data, it seems that the emerging solutions are reacting to the fast pace and evolving nature of the data.
Enterprise Data Characteristics
A useful summary of the current state of the general “data deluge” has been provided by Fox, Hey and Trefethen and is drawn upon here.
We distinguish among three different types of data:
- Observational Data – uncontrolled events happen and we record data about them
- Examples include: astronomy, earth observation, geophysics, medicine, commerce, business intelligence, social data, the internet of things
- Experimental Data – we design controlled events for the purpose of recording data about them
- Examples include: particle physics, photon sources, neutron sources, bioinformatics, engineering design
- Simulation Data – we create a model, simulate something, and record the resulting data
- Examples include: weather & climate, nuclear & fusion energy, high-energy physics, materials, chemistry, biology, fluid dynamics, engineering design
Since most data is yet to be collected, we focus here on data rates rather than absolute amounts. A very high level summary of some of the current or expected data rates in the three data categories is contained in the table below.
Data Type |
Data Rate |
Timing |
Observational | ||
Astronomy: Square Kilometer Array |
>100Tb/sec |
2016-2022 |
Medicine: Imaging |
>1EB/year |
now |
Earth Observation |
4PB/year |
now |
>180PB/year |
now |
|
Experimental | ||
Particle Physics: Large Hadron Collider |
15PB/year |
now |
Photon Sources: Advanced Light Sources |
7TB/hour |
2015 |
Bioinformatics: Human Genome Sequencing |
700Pb/year |
now |
Bioinformatics: Human Genome Sequencing |
10Eb/year |
future |
Simulation | ||
Fusion Energy |
2PB/time step |
now |
Fusion Energy |
200PB/time step |
2020 |
Climate Modeling |
400PB/year |
now |
One immediately notices that the data are hard to compare. The rates for observational data are probably the clearest. For example, if we assume that the Square Kilometer Array were to operate continuously at its full capability, then in the 2022 time frame it would be generating just under 400 exabytes per year. This would appear to make it the world’s largest single data generator – but medical imaging, social data, or the internet of things could well be larger by 2022.
Further note that the business intelligence data used in enterprise computing falls into the category of observational data. This is probably the most difficult data type to deal with. Observational data: is collected continuously; comes from a mix of a small number of large sources (e.g. enterprise data collections) and a large number of smaller – but very significant – sources (e.g. medical imaging, social data, internet of things); and its growth rate increases as the capability to collect and resolve such data increases. So, the associated enterprise computing requirements will be challenging.
Further confirmation of these estimates is provided by the recently launched Chinese Academy of Sciences strategic research project, called NICT (New generation of IC Technology). It assumes that by 2020 the world will need to utilize zettabytes of data. Presumably, most of this data will be observational and it will be used by enterprises.
Convergence
So, if Big Data is HPC and if Enterprise Computing is becoming increasingly dependent on Big Data, will this lead to a convergence of Enterprise Computing and HPC? Judging by the presentations and informal discussions at Big Data’13, such a convergence appears highly likely – if not inevitable – and is, arguably, already underway.
Computing, the internet, social networking, active customer involvement, and the emerging internet of things are causing significant changes in the general business climate and are also creating opportunities for entirely new businesses. A common element in all of this is data. It is coming in large volumes and at high rates. It needs to be analyzed in depth and visualized insightfully to provide useful and actionable business intelligence. As the demands on such intelligence grow and become more complex, their satisfaction will probably require a mix of compute- and data-intensive techniques.
The demands of Enterprise Computing-as-HPC will surely lead to the development of new and interesting applications and algorithms. One can easily conceive of such developments as being similar to what we know from the open literature about Big Data applications in the national intelligence community.
This evolved form of Enterprise Computing may also lead to the development of tailored or special purpose computing systems to support unique requirements. Indeed, the HPC vendor community has already recognized this. One needs only to look to IBM’s Watson or YarcData’s Urika to see the first steps at a response.
Future Opportunities
Big Data technology is currently a young and fragmented market. On the software side, data analytics and visualization are rapidly evolving with dozens of current providers and a steady stream of new entrants with novel approaches. On the hardware side, HPC vendors are adapting or extending current products to the needs of Big Data as well as introducing new products, like Urika and D-Wave‘s quantum computer.
At this point, the data is in the driver’s seat and technologies are reacting to it. There is a lot more data to come. Think of the internet of things (and that subset of it called the internet of us). Speaking of “us”, think of the “crowd” as both a producer and consumer of data – in ever larger quantities and greater varieties. Think also of other emerging data sources, like 3-D printing which, as it matures, will effectively turn material objects into their representations in data.
We are entering a period during which Big Data will transform Enterprise Computing into a principal venue for creative uses of HPC and the incubation of new businesses. The convergence has already started. We in HPC should be full partners in it and help shape the future of Enterprise Computing.