Episode 194: Michael Hunger on Graph Databases

Read vs Write

Graph databases excel in read-heavy applications due to their unique data model. Michael Hunger explains that these databases optimize for reads by precomputing relationships at insertion time, making queries faster compared to relational databases 1. Writes, however, are constrained by disk speed due to transactional requirements, which can limit throughput in write-heavy scenarios. He also discusses the balance between consistency and availability, noting that graph databases can support billions of nodes in a single instance, but consistency across clusters may require eventual consistency setups 2.

Graph databases in general are more supportive in read loads than in write loads, because due to the transactional constraints, each transaction commit has to be flushed to disk.

--- Michael Hunger

This makes them ideal for applications where read performance is critical.

   

Query Efficiency

Graph databases offer significant advantages in query efficiency, particularly for complex queries. Michael Hunger highlights how graph databases simplify recommendation algorithms, which would be complex and costly in relational databases 3. Users can specify patterns in the graph without needing to understand the underlying algorithms, making it easier to derive valuable insights 4. This flexibility allows for the easy addition of new use cases and domains, enhancing the adaptability of applications.

It moves your thinking more around the use cases. So you go away from just thinking about data structures to what you actually want to achieve.

--- Michael Hunger

This shift in approach can transform how developers design and implement applications.