Big Data Requires a Big, New Architecture

“The potential of “big data,” the massive explosion of sources of information from sensors, smart devices, and all other devices connected to the Internet, is probably under-appreciated in terms of its eventual business impact. However, to take maximum advantage of big data, IT is going to have to press the re-start button on its architecture for acquiring and understanding information. IT will need to construct a new way of capturing, organizing and analyzing data, because big data stands no chance of being useful if people attempt to process it using the traditional mechanisms of business intelligence, such as a data warehouses and traditional data-analysis techniques.” Dan Woods; Forbes

So does this apply to Industrial Area, I was heading through Terminal 5 in Heathrow this week, and articles banners around Big Data were all around me, and yes it is the latest “train” for people to board, but is it real in the Industrial Space? As I boarded a train, sat doing a mind thinking moment looking at the industrial operations/ automation landscape I realized why there is confusion is that in the industrial space,  we talk about Enterprise Historians, and one person said to me that is big data! I do not think so, it is just one aspect of the growing industrial information dilemma facing all us over the next 5 years.

When I look at the predictions of Big Data by Industry from Gartner:

The column for “Manufacturing and Natural Resources” which has every row in “Hot” or greater and points to “Volume of data”, “Velocity of data” and especially “Underutilized Dark Data” as Very Hot. This is should not be a surprise to anyone with the historians out there with 10000s of tags soaking up the data at second intervals. In the last 7 years,  Invensys Wonderware has installed 128 million I/O in historian points. Another point not brought out here is the need to make the data “trust worthy” and auditable so business decisions can depend upon it, much of the industrial data is just captured today, not validated against the current state of the process etc.

Now lets understand the “Jobs People want to do today” has there been a change? Yes there has been around the responsibility scope increase. This is both in making decisions and more business impactive decisions, as well as the increase in breadth e.g. Area that a person has to manage.

Initially this seems okay, but  now consider  the devices in the field today, and the amount of data coming from a device that traditionally would have 2 to 3 points, can have 400 points. Is this exaggeration, lets look at an example of a pump.

In the old days,  a pump would have:

• Speed
• Pressure

 Today:

• Speed
• Temperature
• Pressure on incoming and outgoing
• Vibration
• Energy calculations (many variables)
• Number of starts
• Volume
• On goes the list

The reason is that today devices are much smarter this to improve performance, efficiency, maintenance lifetime, and energy consumption management as well as predicting the operational reliability of  the pump. Compared with the old requirement of turning it on and making sure it is pumping to make sure it does not run dry.

Now take one device and put it in a plant context where it is one of 1000s, we have effectively increased the volume of data by 100000s and it will not stop growing. So the ability to capture this data as close to local data source, validating the data, but accessing the data, understanding events, patterns, and relationships across devices, plants, and device types etc required for this ever increasing drive to lower the OPEX costs, through increased efficiency and lower maintenance lower energy consumption  etc.  Again review this data historised for  a pump, the data falls under multiple categories:

·         Operational

·         Energy

·         Maintenance

·         Efficiency

Different roles within the “day to day” running of the industrial operations will analysis the data in different ways, to draw different conclusions. Examples are some people will want to look across multiple pumps and compare efficiency, energy etc vs the Operator who is just look at the current status and availability.

Will the traditional industrial tools be good enough?  I do not think so as all data is not in one form, one data source, take the above time series historian data, combine this alarming, events, and operational data. The introduction of new architectures, “Information Models” and analysis tools which will enable a view across large amounts of data, put this data in the context (this does not mean a data warehouse) and analysis tools quickly bring out trends/ relationships between data from different sources over large areas. All with the simple objective of enable more “real time decisions support”. An example of this is in the latest Wonderware Information Server 2012 R2 (released this month) with a new operational analysis capability. Seen below this capability is out of the box across, MES, Batch, Time series historian data and alarms data sources, providing an immediate view into a trend with a “halo” to show the shift, or batch or phase of operations the process was in, and associated alarm data, all at the operators finger tips.

This is the first step as Invensys will be expanding this capability through the next few years across the Enterprise Control Solution. I will expand on this Big Data in Industry and Decision Support concepts over the next couple of weeks.