While talking about ACID transactions, we have been looking at different approaches to persist data, using Relational, Sharded, and NoSQL/Distributed data stores. They all have one thing in common: They save and retrieve data.
As we compare the different technologies, we tend to think of them as mutually exclusive. Are they really mutually exclusive? Is there a potential flaw in this reasoning?
Let’s take the companies of the early ‘90s, running exclusively on “Big Iron” (mainframe) systems, with proprietary database systems (that look surprisingly like document storage today). We came to the recognition that smaller systems using relational databases, could save a lot of money, and increase performance, or be developed more quickly, allowing us to be more competitive in the market. We didn’t turn off our mainframe. We used two different platforms based on the needs of the customer and the application.
The same principle applies today. We like our structured data warehouse, our wild, untamed data mine, our strictly structured relational OLTP database, and our high performing NoSql counterparts. In fact, there is nothing that keeps an application from using two or more of these different technologies. It all depends on the requirements of the application, and the work to keep each store accurate.
Once again we arrive at the point of a data professional needing to be proficient at Master Data Services. If you’re looking for a new career path, this is one that will continue to grow, in my opinion.
Do you already live in this kind of world? What are the challenges you face? Are there big obstacles you think others should be aware of as they design their systems. Share your thoughts in our comments.
Cheers,
Ben