Methods, systems, and articles of manufacture for scalable processing of collections of files are disclosed. These include retrieving a first file processing job from a job list by a first task, decomposing the first file processing job to sub jobs by the first task when a workload associated with the first file processing job exceeds a threshold, listing the sub-jobs in the job list, monitoring of at least one result storage location by the first task for results from execution of respective ones of the sub-jobs by one or more second tasks, detecting completion of the first processing job based upon the monitoring, and outputting a result for the first file processing job based upon the results from execution of respective ones of the sub-jobs. The job list comprising file processing jobs is populated by a main task.