sdecoret - stock.adobe.com
Alternative databases set for mainstream adoption?
The rise of non-relational databases has been a feature of the data management landscape in the wake of big data – but to complement or replace relational?
With its famous red-coated service staff and reputation for family fun, Butlin’s is a holiday firm steeped in British seaside traditions. But it must also move with the times. Although it takes orders online, the business feared its website performance was not up to scratch.
“Probably the biggest complaint from the businesses was that pages weren’t snappy,” says John Hegarty, head of solution design and delivery, Bourne Leisure, which own Butlin’s, along with Haven Holidays and Warner Leisure Hotels. “People have a certain website dwell time, but any longer and they’re not going to hang about.”
To speed up its customer experience, Bourne Leisure turned to Redis Labs, a distributor of the database of the same name. Redis is one of a group of databases, including GridGain, which offers alternative architecture and improved performance over relational databases. Analysts say they are set to shake up the thinking of senior IT management when it comes to the enterprise database strategy.
Hegarty says he was attracted to Redis because it gave Bourne Leisure the opportunity to boost website performance without replacing the back-end booking engine based on Delphi, a common hotel reservation system based on the Pascal programming language.
“It’s a business rule-based system,” he says. “Every time you go and ask for a price, it executes a whole bunch of business rules, which is incredibly flexible for us. But that does mean it’s slow getting a price.”
As part of a technology refresh, Hegarty’s team built a new caching layer between the booking engine and the website. It serves web users prices from the cache while they search the website. When they come to book, the system draws prices from the reservation engine.
Bourne Leisure already had a caching layer to serve up web prices, based on Microsoft SQL Server, but it did not provide enough performance, says Hegarty.
In 2017, the holiday company started a project to refresh its web technology across all its main businesses and the team took it as an opportunity to look again at the caching database.
“We were on very old technology,” he says. “It was a very ancient Java stack which was really hard to maintain. But because we were that far behind, it was a good opportunity for us to embrace some new ideas.”
Among them was the chance to review NoSQL and in-memory databases. The team opted to work with Redis Labs to build the distributed, in-memory key-value database into the caching system.
This was not the first time the company had looked at in-memory technology. “We’ve had a play with them in the past, but they seemed too complex and too difficult for us to use,” says Hegarty.
After an introduction to Redis Labs and its enterprise Redis database, Hegarty was assured the technology was business-ready and by the offer of 24-hour support. “That ticked the boxes, which some of the other products wouldn’t have because they were just open source,” he says.
“The features we got out of the box were excellent, especially the management features. The support team that looked after our production systems, and installed our application layers, were great on site.”
Read more about new database architectures
- NewSQL database sends NoSQL technology packing at logistics exchange.
- The proliferation of database technologies gives organisations more options to meet data processing needs. However, a strong architecture strategy is a must to avoid a DBMS free-for-all.
- Database security issues shadow distributed data architectures.
The new caching database sits within an application programming interface (API) layer linking it to the front-end website and the back-end booking system. Although the legacy system would not allow the leap to a microservices architecture, the company was able to adopt a “mini-services” approach where reservation API agents can act on behalf of web users. The system is scalable with up to 40 agents operating at one time, says Hegarty.
On Bourne Leisure’s own testing, the new system is a least 10 times faster than the old, although the user experience is more difficult to compare because there was not a like-for-like change in interface. Hegarty says it has been an “incredible success” given the number of new technologies the company adopted in a single leap.
The success is shifting the organisation’s approach to selecting database technologies for future projects. Whereas it would previously default to relational databases, now it would only select them when there is a good reason to do so, he says.
“It has opened up to a whole world of possibilities – it has been an eye-opener for the team,” says Hegarty. “We now have a cloud-first strategy, and with that, where needed, we look for in-memory solutions, or other things that give us that same kind of performance.”
For example, the team is currently considering the cloud-based Snowflake enterprise data warehouse system as an upgrade option, says Hegarty.
Alternatives to relational
Bourne Leisure is not alone in considering alternatives to relational databases. NoSQL and in-memory databases are maturing as viable enterprise technology, says Matt Aslett, vice-president data, AI and analytics, at 451 Research.
“In-memory distributed data layers – including data caching and distributed data grid processing –were initially adopted in industries with the highest latency sensitivity, including financial services and telecoms, as well as internet applications,” he says, “but adoption has spread to other industries, including e-commerce, healthcare and IoT [internet of things].”
Some large enterprises have now standardised on distributed data-grid products as their primary data-processing tier, relegating the database to the role of persistent storage. NoSQL databases were initially deployed to serve applications that were ill-suited to relational databases, but have also matured to support some of the functional requirements, such as transactional consistency, that mean they can be considered potential direct alternatives to relational databases.
Alongside improving the core functionality of their products, they have also taken steps to address enterprise concerns about support, security and scalability, for example, says Aslett.
“We do see emerging databases playing an increasing role for new application development projects and there is considerable growth for new approaches,” he adds.
NoSQL database revenue will grow by 27% a year between 2018 and 2023, while distributed data grid/cache product revenue is expected to grow by 23%, according to 451 Research. By comparison, the relational database market is set to grow at only 6% over the same period.
“However, the relational database continues to dominate the enterprise database landscape, especially given the ongoing reliance on existing applications, and is expected to still account for more than three-quarters of all operational database revenue in 2023,” says Aslett.
Redis, an open source software released under the BSD 3-clause licence, dates back to 2009. It is a key-value database and is unusual in that it provides a data model where user commands do not describe a query to be executed by the database engine, as in relational database management systems, but the operations performed on data types. As a result, data must be stored in a way that is suitable for fast retrieval later.
Howard Ting, chief marketing officer at Redis Labs, says its in-memory features make Redis suitable to the modern application stack. “In the next five years, as organisations continue to move to microservices, storing data in a traditional relational database does not make sense,” he says.
As event-based communications such as microservices find larger adoption, a faster, more responsive database integrated with the cloud is vital, he adds.
GridGain data fabric
GridGain is also an in-memory database, but it differs from Redis. Founded in 2005, it is an open source in-memory data fabric, licensed via the Apache Software Foundation. GridGain is infrastructure software that sits between data obligations and data sources in way that boosts performance and scalability, proponents say.
Terry Erisman, executive vice-president, marketing and alliances, at GridGain Systems, says the approach found a footing in financial services, where Citibank and ING are among the customers, but over the last five years, adoption has spread into other vertical sectors. He says a merchant payment system that took three days to clear running on a mainframe-based relational database took three hours when moved to a new architecture supported by GridGain.
Gridgain can also help with data analytics, says Erisman. “It bridges the gap between the data lake and the operational data store. Combining GridGain with Apache Spark, you can perform analytics on real-time business processes.”
Relational databases still dominate mainstream business applications, but the growing maturity and adoption of alternatives, such as NoSQL and in-memory database technologies, will make IT leaders pause for thought. The move to event-based services architectures is set to prompt a review of the latest options.