Mainframes are used by large organizations as their central transaction processing system. Transaction processing in this context requires high availability, security, performance, and responsiveness.

For example, consumers expect to be able to use their credit card 24 hours a day, 365 days a year. They expect those transactions to be safe and they don't expect to be left standing at the checkout waiting for it to all happen. The mainframe is specifically designed to be the "best of breed" for performing massive concurrent transaction processing in the range of hundreds of transaction per second.

In the examples that follow, we look at typical cases of networks as they are commonly used in high volume business transactions. Each of these examples shows an industry that relies on messages being sent electronically over a communication network. In most cases, a mainframe is used to send the message, one or more mainframes may be needed to route it to the appropriate place, and a third mainframe is used to receive it.

In practice, the number of transactions, the interfaces among the business partners, and the number of data elements is several orders of magnitude more complex.

ATM cash withdrawal

The simple act of withdrawing cash from an automated teller machine (ATM) is much more complicated than it appears. You begin by inserting your identification card and entering a personal identification number (PIN). Your identity is verified online when a computer in the network compares the information you entered to a database of customers belonging to that financial institution. Internal electronic messages are created to access the specific checking or savings account where the money is held. Then, the account balance is verified and approved. Finally, a message is sent back to the ATM to disperse the funds or refuse the transaction.

The withdrawal transaction triggers secondary transactions to update the appropriate checking or savings accounts; this is usually done in real-time. By the time the money is dispensed from the machine, the account balance will reflect the withdrawal. It becomes more complex if you make an out-of-territory withdrawal. For example, you use Bank 1's ATM to withdraw money from your account at Bank 2. The peer bank's database must be accessed and the account status verified.

All of this occurs as the customer waits at the machine. The network and mainframe computers involved must be very fast to keep the response time "reasonable" from the customer's point of view.

The successful completion of the transaction depends on, among other things, both banks using compatible network technology to exchange information.

Credit purchase at a retail store

When you use a credit card to purchase goods from a retailer, then a network, and most likely a mainframe computer, is involved. When your credit card is electronically scanned, the identification is initially handled by the company (Bank) that provides the point of sale credit card reader. From there, the transaction is sent through the network to the credit card company's mainframe. When your account is validated and the transaction is approved, the credit card company issues a debit message to the issuing bank. Concurrently, a credit message to the merchant is issued.

The advantage of sending the transaction immediately is to detect whether you are exceeding your credit limit, and to prevent such violations. Furthermore, if the card is stolen, or if you have exceeded the credit limits, the merchant must be notified in time to void the purchase. Often, an intermediate host is used to handle and approve or disapprove the transaction. All of this can only be effective when a robust, responsive communication network is in place among the merchant, credit card company, and the issuing bank.