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

This whitepaper explores the widespread industry shift of migrating data‑intensive analytical applications from on‑premises to cloud‑native environments, emphasizing that the distinctive cost model of cloud storage demands a more nuanced understanding of performance optimization.

Using observations from Uber’s production deployment of Presto, the authors demonstrate that traditional I/O optimizations often overlook the financial cost of storage API calls, which can lead to unexpectedly high expenses in cloud settings.

The empirical study reveals that Presto’s data‑access patterns at Uber are highly fragmented—over 50% of accesses are smaller than 10 KB and more than 90% are under 1 MB—highlighting a stark contrast with conventional data‑platform environments and underscoring the need for cloud‑specific I/O strategies.

Presented as a case‑study, the paper provides a logical framework and concrete strategies for I/O optimization, aiming to help readers design efficient I/O solutions tailored to cloud environments and significantly improve cost‑performance for data‑intensive workloads.

✓ Adjust cognition and strategies according to different cloud‑storage scenarios and understand their impact on application design and performance.

✓ Examine the additional costs that widely‑used I/O optimization techniques may incur during enterprise‑level cloud migrations, illustrated with Uber data.

✓ Offer a fresh perspective on system design in the cloud‑computing domain to assist stakeholders in addressing the rapid growth of data‑intensive applications.

The whitepaper serves as a starting point for further research, enabling readers to craft specialized I/O strategies that enhance efficiency for data‑intensive applications operating in cloud environments.

For the full content, click “Read the original article” to access the complete whitepaper.