Workload automation is software that runs a company's scheduled computer jobs, mostly overnight, in the right order, and alerts someone when a job fails. A job is one program that runs by itself, without a person, such as loading the day's sales or calculating the stock. A set of such jobs that run together is called a batch. Workload automation is also called a job scheduler or batch scheduler.
Many jobs depend on each other. The stock cannot be calculated before the day's sales are loaded. The purchase orders cannot be sent before the stock is calculated. On one computer, a simple timer can start a program at 11 p.m. But a large company has thousands of jobs on hundreds of servers, in the cloud and in its own data centre. Some still run on a mainframe, a large central computer. Workload automation holds the full list of jobs, and the rule for when each job may start.
Here is a typical case, at a supermarket chain in Hyderabad with 180 stores. At 11 p.m., the first job collects the day's sales from every store's billing system. Workload automation waits until all 180 stores have sent their data. Then it starts the second job, which updates the stock in the ERP. When that finishes at 1:10 a.m., it starts the third job, which calculates how much of each product every store needs. At 3:30 a.m., a fourth job turns the result into purchase orders. Managed file transfer then sends them to 60 suppliers. By 6 a.m., the sales reports are in every store manager's inbox.
One night, store 47 in Secunderabad loses its internet connection, and its sales do not arrive. The scheduler does not start the stock job with incomplete data. It waits 30 minutes, then alerts the IT engineer on duty. The engineer gets the data from the store, and the scheduler restarts the batch from the point where it stopped. The reports arrive at 6:40 a.m., late but correct.
The same tools run month-end work. They start the jobs that prepare the ledger before the finance team closes the books in financial close software. Workload automation suits work that is done in batches at fixed times. Work that must happen within seconds of each change is done differently, as the page on batch and stream processing explains.