Software Architecture

Changing Relationship Types

Changing relationship types is an expensive operation, since each relationship type has to be re-created. When this happens, both nodes connected by the edge have to be visited, the new edge created, and the old edge removed. Hence, edge type or relationship types have to be thought about carefully.

NewSQL Databases

Matthew Aslett first used the term NewSQL to define new databases that aimed to provide the scalability of NoSQL databases while supporting the features of relational databases like ACID. NewSQL databases use different types of storage mechanisms, and all of them support SQL.
NewSQL databases, whose ratings appear in Figure 6-31, improve upon relational databases by providing automated data partitioning or sharding, allowing for horizontal scaling and improved availability, while at the same time allowing an easy transition for developers to use the known paradigm of SQL and ACID.

Figure 6-31. New SQL databases rated for various adoption characteristics

Ease-of-learning curve
Since NewSQL databases are just like relational databases (with SQL interface, added features of horizontal scaling, ACID compliant), the learning curve is much easier. Some of them are available as only Database as a Service (DBaaS), which may make learning them more difficult.
Ease of data modeling
Since NewSQL databases are like relational databases, data modeling is familiar to many and easier to pick up. The extra wrinkle is sharding design, allowing sharded data placement in geographically different locations.
Scalability/throughput
NewSQL databases are designed to support horizontal scaling for distributed systems, allowing for multiple active nodes, unlike relational databases that have only one active leader, and the rest of the nodes are followers. The multiple active nodes allow NewSQL databases to be highly scalable and to have better throughput.
Availability/partition tolerance
Because of the multiple active nodes design, the benefits to availability can be really high with greater partition tolerance. CockroachDB is a popular NewSQL database that survives disk, machine, and data center failures.
Consistency
NewSQL databases support strongly consistent ACID transactions. The data is always consistent, and this allows for relational database users to easily transition to NewSQL databases.
Programming language support, product maturity, SQL support, and community
There are many open source NewSQL databases, so learning them is accessible. Some of the databases also support wire-compatible protocols with existing relational databases, which allows them to replace relational databases without any compatibility problems.
Read/write priority
NewSQL databases are used just like relational databases, with added benefits of indexing and distributing geographically either to improve read performance or write performance.