Bioinformatic tools are often used by researchers through interactive Web interfaces, resulting in a strong demand for computational resources. The tools are of different kind and range from simple, quick tasks, to complex analyses requiring minutes to hours of processing time and often longer than that. Batteries of computational nodes, such as those found in parallel clusters, provide a platform of choice for this application, especially when a relatively large number of concurrent requests is expected. Here, we describe a scheduling architecture operating at the application level, able to distribute jobs over a large number of hierarchically organized nodes. While not contrasting and peacefully living together with low-level scheduling software, the system takes advantage of tools, such as SQL servers, commonly used in Web applications, to produce low latency and performance which compares well and often surpasses that of more traditional, dedicated schedulers. The system provides the basic functionality necessary to node selection, task execution and service management and monitoring, and may combine loosely linked computational resources, such as those located in geographically distinct sites.