Case Study: Oracle Data Cloud achieves significant EC2 cost savings and reduced spot losses with Qubole's heterogeneous cluster solution

Oracle Uses Heterogenous Cluster To Achieve Cost Effectiveness

Oracle Data Cloud, part of Oracle’s data services, runs large, sustained big-data workloads across 82 Qubole-managed clusters. Faced with climbing EC2 bills and high spot-market volatility, the team—running almost entirely on spot instances—suffered catastrophic spot losses, SLA misses and job failures, but could not justify switching most capacity to expensive on‑demand nodes.

To solve this, ODC became the first beta user of Qubole’s heterogeneous-cluster feature, testing on noncritical clusters, tuning configs, then rolling it out across their fleet. The change cut spot losses and substantially lowered costs: on‑demand spend fell from $34K (Sept 2016) to $17K (Jan 2017), the share of on‑demand costs dropped from ~40% to ≤20%, nearly half the clusters are now heterogeneous, and ODC reports large cost savings (up to 90% vs. on‑demand and typically 20–50% vs. homogeneous spot setups) while preserving scale and reliability.

Oracle Data Cloud

Justin Wainwright

System Analysts