# Lack of Graviton usage in Databricks clusters Canonical: https://www.pointfive.co/efficiency-hub/inefficiencies/lack-of-graviton-usage-in-databricks-clusters Databricks supports AWS Graviton-based instances for most workloads, including Spark jobs, data engineering pipelines, and interactive notebooks. By: Matt Weingarten [Cloud Efficiency Hub](https://www.pointfive.co/efficiency-hub) # Lack of Graviton Usage in Databricks Clusters The short version Databricks supports AWS Graviton-based instances for most workloads, including Spark jobs, data engineering pipelines, and interactive notebooks. Matt Weingarten Databricks service [Databricks Clusters](https://www.pointfive.co/efficiency-hub/cloud-services/databricks-clusters) Category [Compute](https://www.pointfive.co/efficiency-hub/service-category/compute) Reference CER-0116 Type Suboptimal Tier or SKU ## Explanation Why the waste happens and who it affects. These instances offer significant cost advantages over traditional x86-based VMs, with comparable or better performance in many cases. When teams default to legacy instance types, they miss an easy opportunity to reduce compute spend. Unless workloads have known compatibility issues or specialized requirements, Graviton should be the default instance family used in Databricks Clusters. ## Billing model The pricing dimensions that drive this cost. Databricks compute is billed based on: - Databricks Units (DBUs): Determined by cluster configuration, including instance type - AWS Infrastructure Charges: Graviton-based instances (e.g., m6g, r6g, c6g) are generally priced lower than their x86 equivalents, billed per second or minute Choosing x86-based instances when Graviton would suffice leads to higher infrastructure costs for equivalent performance. ## How to detect 4 checks to find it in your estate. - Query cluster configurations to identify use of non-Graviton (x86-based) instance types - Check for absence of Graviton enforcement in workspace-level compute policies - Review workload compatibility for clusters running frequently on x86 - Engage with teams to confirm whether Graviton instances have been tested or benchmarked ## How to fix 3 ways to remove the waste. - Monitor utilized instance types and recommend Graviton-based families - Reconfigure default cluster templates to use Graviton by default - Allow exceptions only for workloads with documented compatibility or performance issues ## Documentation Vendor references for pricing and configuration. - [Databricks Support for Graviton Instances databricks.com](https://www.databricks.com/blog/2022/04/18/announcing-databricks-support-for-aws-graviton2-with-up-to-3x-better-price-performance.html) - [AWS Graviton Overview aws.amazon.com](https://aws.amazon.com/ec2/graviton/) ## Related inefficiencies [Browse the library](https://www.pointfive.co/efficiency-hub) - Databricks Clusters CER-0098 ### [On-Demand-Only Configuration for Non-Production Databricks Clusters](https://www.pointfive.co/efficiency-hub/inefficiencies/on-demand-only-configuration-for-non-production-databricks-clusters) In non-production environments - such as development, testing, and experimentation-many teams default to on-demand nodes out of habit or caution. However, Databricks offers built-in support for using spot instances safely. Its job... Compute - Databricks Clusters CER-0159 ### [Inefficient Use of Photon Engine in Databricks Compute](https://www.pointfive.co/efficiency-hub/inefficiencies/inefficient-use-of-photon-engine-in-databricks-compute) Photon is enabled by default on many Databricks compute configurations. While it can accelerate certain SQL and DataFrame operations, its performance benefits are workload-specific and may not justify the increased DBU cost. Many... Compute - Databricks Clusters CER-0022 ### [Missing Auto-Termination Policy for Databricks Clusters](https://www.pointfive.co/efficiency-hub/inefficiencies/missing-auto-termination-policy-for-databricks-clusters) In many environments, users launch Databricks clusters for development or analysis and forget to shut them down after use. When no auto-termination policy is configured, these clusters remain active indefinitely, incurring unnecessary... Compute --- Source: the public page above. Product screenshots and illustrative interfaces are examples, not live customer data.