What are the criteria for successful application of large data?

It's certainly the right thing to do in large data categories, and start-ups and corporate giants around the world are also creating value by leveraging large data and large data applications-turning large amounts of data into money or competitive advantage. However, behind the glory, always conceal some of the truth can not be ignored. In short, not all attempts at big data have been rewarded, and far from it. There is another truth to be reckoned with, in the IT business world, where the standard of "success" of large data is loosely defined, and even the notion that "we do not fail completely" can be attributed to "success".

So what are the criteria for the success of large data applications? 10gen Strategy Vice president Matt Asay brought him 4 criteria for success:

First, it must be operational.

Big data should create real value for the industry, not just high-tech. In a report on the future of big data, McKinsey points out that big data has trillions of potential value in healthcare, government, retail and manufacturing industries. The organization's successful implementation of large data needs to bring tangible rewards in a few ways: additional revenue, improved customer satisfaction, cost reduction, and so on.

Secondly, there must be an essential improvement

Large data delivery should not only be incremental business model improvement, but also a fundamental breakthrough. For example, in the case of start-up Foursquare, in order to discover the relationship between data, Foursquare used machine learning algorithms to enable the system to build "Explore", a social referral system can provide users with valuable location information in real time, Use the new business model to drive the location information type business. "Explore" relies on large data technologies while gaining insights from more than 30 million location information. Now Foursquare has the ability to understand how people interact, and location information is not just a platform but a real world.

Again, must have a high speed

Traditional database technology can pull down the performance of large data, but also very cumbersome, because regardless of whether the technology to meet your needs, patent licensing involved in the enterprise cumbersome system far beyond your imagination. A successful large data project, the use of toolset and database technology must meet both the volume of data and the dual needs of diversity. The argument is that a Hadoop cluster can be built in just a few hours and can provide fast data analysis when it is finished. In fact, most of the big data technologies are open source, which means you can add support and services to your needs, and licensing is no longer a hindrance to rapid deployment.

Finally, you must be able to

Before the big data, similar gilt groupe such "time-snapping" companies could not be achieved. Timed-snapping sites need to be processed by tens of millions of users on a daily scale, and can cause very high server load spikes-making this business model possible with high-performance, fast-expanding large data technologies.

Summary

The key to success in large data deployments is not how much data the system can handle per second, but how much value it brings to the company's business after using large data and whether there is a breakthrough in the business. Focus on the type of business and choose a tool set that is appropriate for your company's business.