Toy Hadoop is a distributed compute engine that improves upon Hadoop's ability to handle metadata-intensive workloads with many small files. It uses FusionFS as the underlying distributed file system due to FusionFS's efficient metadata management. Toy Hadoop addresses Hadoop's "small files problem" by grouping small files together and assigning them to task trackers in batches, rather than processing each file individually. This reduces communication overhead and improves performance even for small file workloads. Evaluation shows that Toy Hadoop has significantly better performance than Hadoop for files under 100MB, especially when using FusionFS for its metadata handling capabilities.