Rethinking Cloud Telemetry for Performance and Scale

Published February 11, 2026

Modern cloud services rely on telemetry—the collection, transmission, and analysis of data—to monitor and ensure that systems are healthy, performant and reliable. But gathering and analyzing this data at scale is often expensive and slow, forcing engineers to balance trade-offs between performance, cost and accuracy.

Zeying Zhu, a fifth-year computer science doctoral student at the University of Maryland, is working to overcome these limitations.

“Modern telemetry systems should be high-performance, scalable and inexpensive, but today’s solutions struggle to achieve all three at once,” she says.

Zhu’s research applies approximation techniques to make cloud monitoring more efficient, producing results that are close enough to exact answers while using far fewer resources.

Working in the lab of her adviser, Alan Liu, an assistant professor of computer science with an appointment in the University of Maryland Institute for Advanced Computer Studies (UMIACS), Zhu develops large-scale, high-performance, and resource-efficient data systems.

One recent project, PromSketch, allows telemetry queries to run up to two orders of magnitude faster by reusing approximate summaries of past data rather than repeatedly scanning enormous databases. Another, NetMigrate, addresses challenges in migrating data across cloud storage systems during shifting workloads, using lightweight tracking to maintain performance and reliability.

Liu says Zhu’s work stands out for both its rigor and real-world relevance.

“What sets Zeying apart is her persistence and clarity of thinking,” he says. “Her research in cloud observability and data systems tackles timely, fundamental challenges in understanding and operating large-scale cloud systems, and it’s work that both researchers and practitioners can learn from.”

For Zhu, the appeal lies in finding solutions that improve performance without driving up costs.

“These projects let us explore how to make cloud telemetry not just faster, but also more cost-effective,” she says. “Finding elegant solutions that reduce operational costs and latency has always fascinated me.”

Looking ahead, she is exploring ways to reduce the cost of collecting telemetry data itself by applying approximation techniques directly at data sources, cutting down how much redundant data needs to be transmitted and stored.

Zhu has also recently begun collaborating with Ian Miers, an assistant professor of computer science with an appointment in UMIACS, on verifiable network telemetry—a project that allows internet service providers to prove the correctness of their analytics without revealing sensitive network information. Such systems are important for enforcing regulations like network neutrality and verifying service-level agreements.

Both Liu and Miers are core members in the Maryland Cybersecurity Center.

Collaborating with Miers has expanded Zhu’s perspective on privacy-preserving systems and their real-world impact.

“Working with Ian, I’ve learned how advanced zero-knowledge proof techniques can be applied to real-world networking systems,” she says. “It’s a unique challenge to balance strong privacy and security with fast performance.”

Zhu’s interest in computer science began in high school in China, where she competed in algorithm competitions and explored various algorithms balancing between memory and computational efficiency. That early fascination led her to study system research of networking at Xi’an Jiaotong University, and eventually to her current focus on cloud telemetry, where efficient algorithms and system design go hand in hand to reduce cost and latency at scale.

Outside the lab, Zhu enjoys running, basketball and badminton. She recently completed her first 10-mile race, running for enjoyment and scenery rather than competition.

After completing her Ph.D., Zhu plans to pursue a faculty position, continuing to explore ways to make large-scale computing systems more performant and more resource-efficient.

—Story by Melissa Brachfeld, UMIACS communications group