Full Speed Ahead
Practical tips on using NoSQL to accelerate real-time insights in a legacy environment.
In our last post, we discussed why and how you can maintain legacy systems — without foregoing modern BI capabilities. NoSQL offers a powerful way to monitor KPIs in close to real-time, which helps with decision-making, forecasting, and SLA adherence.
Under a legacy approach to BI, data from e-commerce transactions is likely to languish in data warehouses. But with a NoSQL database in place, that same information could instead be accessed for up-to-the-second insights into consumer behavior. For example, 60 percent of businesses do not use their websites to create and sustain customer loyalty, in large part due to ineffective system integration. NoSQL platforms give organizations a way to work around this issue.
Consider the case of Marriott, which has already implemented NoSQL with great success. The hotel chain — with more than 4,200 properties in 79 countries — recently moved its central reservation system to NoSQL. According to Diginomica, the main driver behind the change was the need to better compete in the digital economy by using an open architecture that enabled faster, more reliable deployment of critical applications. Marriott gained high availability, scalability (aided by commodity hardware) and overall superior business agility through its switch to NoSQL. Once Marriott’s properties are fully live, there could be 30 million documents being accessed at 4,000 transactions per second.
Only NoSQL can provide the efficiency necessary to operate at such scale. Because it has matured so much, companies still maintaining their legacy mainframes can now begin thinking about how to modernize their databases to spur and sustain growth.
Practical tips: Implementing NoSQL into legacy BI infrastructure
NoSQL is a perfect option for organizations that cannot fully replace existing BI systems based on relational databases and data warehousing. For those planning to integrate NoSQL, it is important to know when and where to take advantage of its distinctive capabilities:
- Make sure to keep transactional data in a relational database for regulatory compliance and simplification. This also guarantees integrity, atomicity, consistency, isolation and durability for database transactions, through the reliability of a familiar legacy system.
- However, if you currently use a commercial data warehouse from an online transactional processing vendor, think about augmenting and eventually replacing it with a NoSQL database. NoSQL delivers the scalability and low overhead you need for advanced applications such as e-commerce.
- More specifically, key-value stores (for storing user sessions and shopping cart information), graph stores (for roads, maps, social links, recommendations), document stores (for a product catalog or completed orders) and column families (large scale web analytics and support for reports) can all be utilized through NoSQL.
Identifying and enforcing these requirements from the start will make the transition to NoSQL as smooth as possible. Organizations can make big changes that will benefit their business without ripping and replacing their entire BI systems.
There are a few pointers to making NoSQL work well:
- First, ensure there is a solid architecture in place to enable the seamless integration of NoSQL and in-memory computation systems. If a legacy setup does not allow such compatibility, it may not be possible to use your NoSQL database to its fullest capacity.
- Second, identify integration points in the legacy application stack, as integrations can happen at the Object Relational Mapping or database layer without any additional overhead.
- Third, the NoSQL database must be able to interact with the OLTP database to obtain the best information for BI projects. OLTP is used as a relational database. NoSQL, on the other hand, is geared more toward analytics, since it is more responsive to changes in attributes. Accordingly, its ability to perform rapid data insertions can help organizations keep up with shifting user demands and get the best information in near real-time.
Scale, accelerate, optimize: Tapping the real benefits of NoSQL
Once your NoSQL databases are set up and configured, you can get an early advantage by eliminating unnecessary legacy processes. For example, data warehousing requires high-overhead extract, transform, load (ETL) operations to read and enrich data from the OLTP. By contrast, NoSQL databases do not rely on ETL.
Another benefit is easy scalability. NoSQL databases excel at locating specific items in vast data sets, from a wide variety of sources. Compared to relational databases, they are optimized for horizontal rather than vertical scaling. Organizations can simply add more servers as they expand, rather concentrating more capacity in an existing system.
At a technical level, NoSQL uses in-memory computation to store data and perform complex calculations. It removes reliance on separate memory cache layers and allows quick access to frequently used information. Overall, NoSQL offers high throughput and low latency for big data workloads, all without erasing the functionalities and interconnections of legacy BI tools.
Looking ahead, many more companies will implement NoSQL databases, in some cases alongside relational ones. The increasing speed of decision-making and volume of data to process mean many organizations must travel the path to real-time analytics.
We’d love to hear about your experiences on that journey.