Abstract: The long-term persistence of forest ecosystems hinges on their resilience to ongoing disturbance. Quantification of resilience in these valuable ecosystems remains difficult due to their vast extent and the longevity of forest species. Resilience to wildfire may arise from feedback between fire behavior and vegetation structure, which dictates fuel loading and continuity. Regular fire generates structural variability which may then enable forests to withstand future fires and retain their fundamental properties and functions-- a hallmark of a resilient system. A century of fire suppression in the western United States has homogenized the structure of many forests, potentially upsetting these feedbacks and compromising forest resilience. We investigate the generality and scale of the effect of structural variability on wildfire behavior in yellow pine/mixed-conifer forest of California's Sierra Nevada using cloud computing and texture analysis of a 33-year time series of satellite imagery. We measure wildfire response to forest structure for an unprecedented number and size range of wildfires, ensuring representation of both typical and extreme fire behavior, and find that greater structural variability is strongly associated with a lower probability of fire-induced overstory tree mortality. This resistance to wildfire was most apparent at the smallest spatial extent of forest structure tested (90m x 90m). Local-scale structural variability thus links past and future fire behavior, and makes forests more resilient to wildfire disturbance. Management strategies that increase vegetation structural variability, such as allowing fires to burn under moderate fuel and weather conditions, may therefore increase the probability of long-term forest persistence. Significance: A "resilient" forest endures disturbance and is likely to persist. Resilience to wildfire may derive from variability in vegetation structure, which interrupts fuel continuity and prevents fire from killing overstory trees. Testing the generality and scale of this phenomenon is challenging because forests are vast, long-lived ecosystems. We develop a novel cloud computing approach to consistently quantify forest structural variability and fire severity across >30 years and nearly 1,000 wildfires in California's Sierra Nevada. We find that greater small-scale structural variability increases resilience by reducing rates of fire-induced tree mortality. Resilience of these forests is likely compromised by structural homogenization from a century of fire suppression, but may be restored with management that increases structural variability of vegetation.