Building performance research using various informatics techniques has progressed extensively in the last twenty years by advancing the fields of automated fault detection and diagnostics (AFDD), commissioning, data mining, and visualization for commercial buildings. Despite this effort, it has been difficult to understand the effectiveness of different approaches as compared to each other as there is a lack of general, public benchmarking datasets for this industry. We propose a repository in which researchers can release their detailed raw datasets for the purpose of repeatability, benchmarking, and utilization by other researchers. We start this effort through the public release of a single, large building performance seed dataset. The dataset is from a primary and secondary school campus that has 76,000 square meters floor area of conditioned, indoor space in seven buildings that include classroom, office, sports facilities, auditorium, cafeteria, dormitory, and mixed-use spaces. The dataset contains almost 3 years of detailed temporal data from 3,690 measured data points, most of which are sampled at a frequency of 1-3 minutes. The campus is located in a tropical climate with a continuously high cooling and dehumidification load. Some of the dataset has been annotated with building event schedules and known anomalous behavior which can be used as ground truth for detection algorithms. The dataset is available for download online and will serve as the first example in a planned repository of raw datasets from various buildings from different climates and contexts.