Supporting an Analytics Strategy in a Hybrid Cloud
The lure of an analytics strategy in the hybrid cloud is cloud elasticity. Your data can be processed across clusters of computers. This means that the analysis is occurring across machines. If you need more compute power, you can get it from the cloud.
Big data analytics in a hybrid cloud
Here are some examples of where analytics get big and may require cloud resources:
Financial services: Imagine using advanced analytics technologies like predictive analytics to analyze millions of credit card transactions to determine whether they might be fraudulent. Or, on the unstructured side, picture the text in insurance claims being analyzed to determine what might constitute fraud.
For example, take a worker’s compensation claim submitted by a worker who may have been reprimanded several times by his boss. This data (or the claim), which came from unstructured sources, can be utilized together with structured data to train an analytical system on what patterns might indicate fraud. As new claims come in, the system can automatically kick out the ones that may need to be investigated.
Retail: Just think about the recommendation engines from Amazon and eBay. They’re becoming more sophisticated. eBay is using advanced technologies that will look at what you’re purchasing and then, based on models it has of the numerous purchases of other people, make a recommendation.
Another example is the use of advanced analytics over massive amounts of data in real time at big-box stores. Using your loyalty card, based on what you’re buying, what you have bought in the past, and what others with similar profiles like you have bought, the store will provide you with coupons for different products you might like.
Social media analysis: Imagine all the data being collected across the Internet. This includes blogs, tweets, and newsfeeds. Companies are mining this unstructured data to understand what is being said about them. For example, a consumer packaged goods (CPG) company might mine this data to determine what is being said about them and whether this sentiment is positive or negative. Numerous companies are providing this kind of service in the cloud.
Writing the code to process this data across clusters of machines requires highly trained developers and complex job coordination. With a technology like MapReduce, the same MapReduce job that is developed to run on a single node can distribute this analytic processing power to a group of 1,000 nodes. Say you need immediate analysis of sensor data or social media data that is streaming into your data center or your cloud provider. Parallel processing across multiple computing resources can help to do this by spreading the analysis across the environment. It gets you to insight faster.
Other cloud analytics
The cloud can be useful in supporting an analytics strategy when your data isn’t that big (in contrast to the previous example of big data). Say you work at a company that wants to predict what action your customers will take. You want to use predictive analytics to do this, but you don’t have the skills in-house. In this case, you can turn to analytics providers that offer SaaS-based services for help. You provide them your data, and they provide you with the analysis.
A number of cloud-based offerings on the market can either help you analyze your data or provide software in the cloud for you to do the analysis yourself. Maybe you’re using a cloud-based CRM and ERP system, and you want to analyze the data that’s being generated there. There’s a cloud service for that.