Praval Panwar, Microsoft
SRE teams are often told the cloud is “infinitely scalable”—until costs spike or systems fall over. In practice, many teams oscillate between over-provisioning and panic-scaling, reacting to incidents instead of planning for uncertainty.
Other industries like airlines, power grids, and logistics have faced this problem for decades. They don’t assume infinite resources. They forecast demand, build buffers, and design explicitly for failure. This talk borrows those proven ideas and applies them to cloud systems. It explores why assumptions of infinite scale lead to wasted spend and fragile reliability, and how SRE teams can reason more clearly about capacity, cost, and performance together. The focus is not on new tools, but on better mental models—so teams can plan for spikes, failures, and uncertainty without relying on guesswork or excess capacity.
Praval Panwar is a Principal Software Engineer at Microsoft, where he works on large-scale observability and telemetry platforms. He has spent over seven years designing and operating hype-scale distributed systems that need to work reliably under messy, real-world conditions. His interests include how systems fail under pressure, how assumptions break at scale, and why many operational problems turn out to be design problems in disguise.
author = {Praval Panwar},
title = {Infinity Is Not a Strategy: {Right-Sizing} the Cloud},
year = {2026},
address = {Seattle, WA},
publisher = {USENIX Association},
month = mar
}
