Joyce Kay Avila - Snowflake: The Definitive Guide

The shared-disk architecture was an early scaling approach designed to keep data stored in a central storage location, accessible from multiple database cluster nodes (). The data accessed by each of the cluster nodes is consistently available because all data modifications are written to the shared disk. This architecture is a traditional database design and is known for the simplicity of its data management. While the approach is simple in theory, it requires complex on-disk locking mechanisms to ensure data consistency. Additionally, concurrency is a major problem and adding more compute nodes only compounds the problem. Therefore, the true scalability of this architecture is limited.

Figure 1-1. Shared-Disk Architecture

The shared-nothing architecture, in which processing and compute is scaled together (Figure 2-2), was designed in response to the bottleneck created by the shared-disk architecture. This evolution in architecture was made possible because storage had become relatively inexpensive. However, distributed cluster nodes along with the associated disk storage, CPU, and memory, requires data to be shuffled between nodes, which adds overhead. Depending on how the data is distributed across the nodes will determine the extent of the additional overhead. Striking the right balance between storage and compute is especially difficult. Even when it is possible to resize a cluster, it takes time to do so. Thus, organizations often overprovision shared-nothing resources, which results in unused, unneeded resources.

Figure 1-2. Shared-Nothing Architecture

Most NOSQL solutions rely on shared-nothing architecture; thus, they have many of the same limitations. However, the benefit of NoSQL solutions is that they can store non-relational data without first requiring transformation of the data. Additionally, most NoSQL systems dont require schemas. NoSQL, a term that implies Not Only SQL rather than NO to SQL, is a good choice for storing e-mail, web links, social media posts and tweets, road maps, and spatial data.

There are four types of NoSQL databases. Document-based NoSQL databases such as MongoDB store data in JSON objects where each document has key-value pair like structures. Key-value databases such as DynamoDB are especially useful for capturing customer behavior in a specific session. Cassandra is an example of a column-based database where large numbers of dynamic columns are logically grouped into column families. Graph-based databases, such as Neo4j and Amazon Neptune, work well for recommendation engines and social networks where they are able to help find patterns or relationships among data points.