Infrastructure Efficiency and Optimization

3 posts · Jun 2020 – Aug 2026

What this trail covers

Dropbox's ongoing efforts to improve infrastructure efficiency, focusing on planning, continuous optimization of the active fleet (including Deep Sleep and workload balancing), and extending hardware lifecycle. This capability evolves to address growing demand, particularly from AI workloads, by maximizing resource utilization and minimizing waste. This post details how Dropbox reduced its monorepo size by 77% (from 87GB to 20GB) by addressing Git's delta compression heuristics, significantly improving developer velocity and mitigating operational risk associated with repository size limits.

All Dropbox trails

Aug 2026Improving infrastructure efficiency for growing demand in the age of AIThis post details Dropbox's system-level approach to infrastructure efficiency, focusing on planning, continuous optimization of the active fleet (Deep Sleep, workload balancing), and extending hardware lifecycle. It emphasizes understanding tradeoffs between different infrastructure layers (e.g., storage density vs. energy/cooling) and optimizing for metrics like watts per petabyte. The post highlights how these efforts enable growth and accommodate AI workloads by maximizing resource utilization before expanding footprint.
Mar 2026Reducing our monorepo size to improve developer velocityThis post details how Dropbox reduced its monorepo size by 77% (from 87GB to 20GB) by identifying and addressing an issue with Git's default delta compression heuristic, which was incorrectly pairing files from different languages due to path similarities. The solution involved a more aggressive repack using tuned window and depth parameters, executed on GitHub's servers, which reduced clone times from over an hour to under 15 minutes and mitigated the risk of hitting GitHub's repository size limit. The rollout was treated as a production infrastructure change, involving testing on a mirror and monitoring key metrics.
Aug 2021How we’re making Dropbox data centers 100% carbon neutralDropbox has achieved 100% carbon neutrality for its data center storage server power by implementing three main strategies: maximizing power usage effectiveness (PUE) through outside air economization, thermal containment solutions, and maximizing power utilization; optimizing overall power consumption by quickly powering down decommissioned hosts, leveraging HDD Standby for idle servers, rightsizing capacity through monthly planning and system monitoring, and moving orchestration to Kubernetes for improved resource utilization (bin packing, oversubscription, background jobs).
Jun 2020Speeding up a Git monorepo at Dropbox with <200 lines of codeThis post details how Dropbox addressed Git performance degradation in its large monorepo, particularly on macOS. The solution involved a combination of upstream Git improvements (fsmonitor, untracked cache, split-index mode) and a small wrapper script (<200 lines of code) to automatically configure Git for developers. This resulted in significant reductions in p50 and p90 durations for common Git operations like `git status`, without fragmenting the repository or requiring additional services.
Advertisement