Tomicevic posits that relying on “brute-force compute” by easily spinning up hundreds of cloud servers can be a form of architectural laziness that wastes both power and financial resources. He suggests that the responsibility is increasingly on engineers to consider the sustainability impact of their work. This perspective challenges developers to move beyond default solutions and explore more efficient methods for modeling and solving complex problems.
Contrary to some criticisms, Tomicevic asserts that cloud providers are strongly incentivized to optimize energy efficiency, as their profit margins depend on it. He argues that consolidating workloads in the cloud is often more carbon-efficient than maintaining underutilized on-premises data centers. Regarding AI-generated code, he cautions that it is not yet a guaranteed path to efficiency, citing research that suggests it can be slower and more memory-intensive than human-written code
in some scenarios, leading to higher energy consumption per task.
The core of Tomicevic’s solution is a shift in data modeling. He advocates for using graph databases for a wide range of problems where data is highly interconnected. He explains that for use cases like recommendation engines, fraud detection, and route optimization, traversing a native graph is far more efficient than the resource-intensive JOIN operations required in traditional relational databases. Recomputing complex joins in a relational database can be far more resource-intensive than traversing a native graph,
he states, positioning graph models as a key tool for achieving the goals of both FinOps and GreenOps.
Tomicevic calls for more creative thinking from developers and architects, urging them not to default to the relational approaches commonly taught. The next step for organizations is to evaluate their cloud workloads to identify opportunities where a graph model could lead to significant efficiency gains. This involves a trade-off between immediate business agility and long-term architectural optimization.
Based on the arguments, developers and IT leaders should consider the following actions:
- Analyze existing applications for relationship-heavy queries that may be better suited for a graph model.
- Evaluate graph databases like Memgraph for new projects involving complex, interconnected data.
- Scrutinize the performance and resource consumption of AI-generated code rather than assuming its efficiency.
- Incorporate sustainability and cost-efficiency metrics into architectural design and review processes.
Follow Hashlytics on Bluesky, LinkedIn , Telegram and X to Get Instant Updates