Big Data and cloud: far from perfect

Big data is fashionable and promises a lot of things, but when organizations try to consolidate their data into large databases in the public cloud, they ignore some technical realities.

These days, big figures seem fashionable. It's really big, it's new, it's about Hadoop, and it's generally in the public cloud. Now every week new businesses and new cloud products appear, promising to eventually control the company's data. They are all peddling the same idea: companies can migrate to almost "no capacity limit" PB-level databases through the resilience of the public cloud.

And the reality and such speculation is far from the difference. When organizations try to consolidate the enterprise's data into large databases in the public cloud, they ignore some technical realities.

First, big data means the challenge of big migrations. Getting data from the enterprise to the public cloud may be problematic. Although the enterprise every day to carry hundreds of thousands of data records on the open Internet, but in many large data situations, we are talking about millions of data records how to move from the existing enterprise system, conversion, encryption problem.

In doing so, you will soon encounter bandwidth constraints. In fact, many companies are using the http://www.aliyun.com/zixun/aggregation/31488.html "> FedEx Flash drive to their public cloud provider."

Second, although security measures are sure to be implemented in the cloud, it is generally much cheaper to handle data levels in a system or private cloud within an enterprise. In many cases, the enterprise security model and security technology costs are not very high. For example, in a public cloud, an enterprise must encrypt its data, but not in its own data center. Similarly, keeping data local, compliance is generally easier and less costly.

By the way, I'm not saying that big data doesn't fit the public cloud at all, but that you have to think about all the technical issues. Plus other architectural issues, you have to specifically analyze the problem

(Responsible editor: Duqing first)