Big data solutions will mature over the next two-three years and this maturity will be accompanied with falling prices. That’s the good news for IT this holiday season. The other key for IT is that it needs to put its arms around big data now-by getting big data roadmaps together and understanding how big data will play in the data center.
Just how do you build a big data roadmap?
Looking at online gaming technology isn’t a bad place to start, as it’s likely that the big data technologies emerging in online gaming will be incorporated into enterprise big data solutions.
Let’s take a look.
Graphics Processing Unit (GPU) Cards-Today they render battle wounds “real,” optimize falling buildings and create realistic and action-packed mid-space collisions-but tomorrow GPUs could bring this same sense of immediacy and “instant action” to big data. GPUs are mini-parallel processors that blast through numerical algorithms and probabilities for real-time game simulations-and there is no reason this parallel processing can’t be harnessed to collect and analyze data for social media analysis, predictive modeling or even Internet of Things machine to machine information collection and control.
Michael Upchurch, Chief Operating Officer of Fuzzy Log, a database analytics company, believes that GPU technology will be instrumental in transitioning big data applications that were once strictly the province of scientists, researchers and IT specialists to business users. He points to a call center customer use case where GPUs speed real-time analytics processing that in turn assists call center efficiency and eliminates call center caller frustration by correctly and rapidly determining which call center rep is best suited to handle a particular call.
New noSQL Database Technologies-The third and fourth generation report generators of the past, and many of the business analytics of today, utilize relational databases that were first conceived in the 1970s. These databases, constructed in rows and columns, gave significantly greater latitude for searches than many of the hierarchical databases that transactions were stored in. But relational databases also operate on structured data. If an analytics analyst wants to look at an unusual combination of events for business analysis (e.g., “How many men under 40 in Seattle go fishing, actively promote fishing on Facebook to their friends and purchase at least $1,000 worth of gear each year?”), pieces of different relational databases most likely will need to be joined by a database administrator into a datamart that can support the analysis. Once the database joins are done, there is still no guarantee that data retrieval and performance will be fast enough to meet the needs of the business-and if the query changes, all of the initial join work might have to be done again.
The noSQL databases coming into the market now ease the situation for analytics considerably. First, they don’t require database joins, so DBAs have to do less work. Second, they are capable of storing structured, unstructured and semi-structured data, so they have much more range than relational databases. Third, these databases are capable of accepting new data at any time (whether or not it is structured). This makes them uniquely malleable for business analytics.
Knowledge Engineering-Online gaming is a serious entertainment business that must not only attract new players, but also keep existing players engaged. To “keep the product interesting,” gaming companies like Riot Games build staffs of business analysts who relentlessly study user behavior. Riot Games averages 32 million users per month, and has a dedicated BI (business intelligence) team that asks questions like which game champions and skins (costumes) are most popular in certain geographical areas?
What are the takeaways for IT?
IT should look at the online gaming industry with the intent of identifying best practices and big data goals for IT roadmaps. Databases and data warehouses are certainly going to behave differently with big data and DBAs should be researching this now. Within the data center infrastructure, we can expect to see continuing convergence of devices and technologies that will include GPU integration. This is the time that network, systems and applications specialists should be looking at this convergence and what it ultimately portends for the data center. Finally, there is knowledge engineering itself–an emerging discipline in enterprise IT and end business areas, that won’t be in an “emerging” state for long. CIOS need a staff development (or acquisition) plan that keeps pace with their plans for big data technology adoption.