Optimizing I/O for Data‑Intensive Analytics in Cloud‑Native Environments: Insights from Uber Presto

This article explores the widespread industry trend of migrating data‑intensive analytics applications from on‑premises to cloud‑native environments. It highlights that the unique cost model of cloud storage requires a more detailed understanding of performance optimization.

The paper presents an empirical study of Uber's production Presto deployment, showing that traditional I/O optimizations often ignore the financial cost of storage API calls, which can lead to high expenses in the cloud.

Observations reveal that over 50% of data accesses are smaller than 10 KB and more than 90% are under 1 MB, indicating a highly fragmented access pattern that has different implications in cloud settings compared to traditional data platforms.

Through a case‑study approach, the authors provide logical I/O optimization strategies designed for cloud environments, aiming to help readers design efficient I/O solutions that significantly improve the cost‑performance ratio of data processing in the cloud.

The whitepaper offers a fresh perspective on system design in the cloud computing domain, assisting stakeholders in addressing the rapid growth of data‑intensive applications.