It’s the scale-out architecture, not NoSQL
New applications are now primarily being developed for the cloud environment – whether it is private, public, or hybrid. The data layer for a modern application is complex with multiple data stores, and choosing the correct database requires understanding a market and database landscape that is complex, fragmented, and rapidly evolving.
NoSQL is one way to characterize the new crop of data stores, but the term misses the mark by diverting focus to the wrong aspect. While NoSQL means “not only SQL”, it has come to symbolize the absence of SQL and new proprietary query languages that are often simplistic.
The better question is: what is the primary value I get from these data stores? Furthermore, is there value in having a non-standard and less expressive language than SQL?
The real thread binding these new data stores is their scale-out architecture and the ability to incrementally and predictably scale on commodity hardware. The move to the cloud assumes the ability to provision generic hardware quickly and the flexibility to grow seamlessly with increasing workloads.
The legacy data layer has just not been able to do this; for example one does not have the option of putting big-iron Oracle boxes in the public cloud. Those applications that can use NoSQL stores increasingly have chosen to do so. Many have even moved to the public cloud. But those with the legacy data layers have more or less been required to stay on premise.
The inexorable trend toward expressiveness and analytics
Databases have guarantees and rich query languages primarily because applications need them. You can start with a simple NoSQL database that does simple reads and writes, but as your business grows there is a need to better understand what is going on with the business and therefore run analytics. Once there is revenue associated with interactions, the transactions and ACID properties get more important.
The scale-out NoSQL data stores have provided scale and speed, and are now adding more features. Simple query languages are giving way to more complex ones, with features such as aggregation frameworks and schema enforcement being added.
We see many of the SQL features are being re-invented sans the join operator. Because normalized schemas and joins are the workhorses for OLTP, joins need to be part of the architecture and are not so easy to patch on later. This is where NoSQL primary databases hit a limit. Therefore it is not surprising there is growing interest in scale-out SQL (NewSQL) databases that promise to continue this march toward more features.
Clustrix: The most feature-rich scale-out data store
Today when you’re running a new application that is engineered for the cloud and using multiple scale-out data stores, how do you know which database is the right option?
The answer depends on the application, most of which require multiple data stores. Chances are that a relational database is a good fit for a large component of your application, alongside NoSQL stores.
Among the scale-out SQL (NewSQL) databases, most are focused on transaction scaling. Clustrix scales transactions, but also provides fast analytics and reporting on your primary database. Clustrix has reached MySQL feature set and will continue further development of robust analytics. This gives you an alternative that can replace complex existing workloads not just the simplistic ones. We have many customers who have websites serving millions of users as well as those who have used Clustrix to get faster real-time reporting and analytics.
With 20 commodity servers, you can build a 320-core database in minutes that has enough horsepower to match a scale-up monster. This scale can be doubled seamlessly by simply adding more nodes. This database will allow you to run transactions and analytics both. Scale-out SQL databases have proved themselves in transactional workloads and dominate offline analytics alongside Hadoop with examples such as Vertica, Greenplum and Redshift. Recently, there is resurgence of interest in the value provided by SQL and we believe the scale-out SQL databases will be one of the mainstays of databases in coming years.