The mainframe architecture was the only physical computer architecture that was available in the 1960s, 1970s and 1980s. With the procurement of a mainframe computer at least several millions of Euros were involved, and as a result only large industrial companies, financial institutions, governments and universities could afford the use of these computers. IBM was the undisputed market leader in mainframe computing.
Mainframe architecture is characterized by maximal centralization of the elements of the architecture, with minimal decentralized elements and limited network traffic. The mainframe itself is a large central computer that is based in the data centre of the organization. The mainframe houses all disk space, internal memory and processing power. The only architectural elements that are installed decent rally, on the locations where the users of the system carry out their work, are data entry and output devices. These devices are so-called dumb terminals, simple monitors without graphical capabilities, and keyboards. The terminals have no processing power, no disk space, and very limited internal memory which is just sufficient to display characters on the screen. In the early days of mainframes even printers were only available centrally, in order to minimize network traffic. Prints were treated like traditional mail: they were delivered once a day to the pigeon-holes of the users.
When an ERP system is installed on physical mainframe architecture, the principle of centralization has to be followed. The ERP database is installed on the disk space of the central mainframe, as well as the software in which the business logic has been programmed. The calculations and data manipulations in the ERP business logic are carried out by the centralized internal memory and the central processing unit. The only decentralized architectural element is the interaction, which is present at the site of the individual users with their dumb terminals; all other elements are installed centrally and are shared by all users. In Figure 4.2 a schematic drawing of the elements of mainframe architecture is presented. At the bottom of the picture, the elements of the physical architecture are named, as well as the mapping of the logical ERP elements on the elements of the physical architecture.
The main advantage of centralized mainframe architecture for ERP is the ease with which the database and the business logic can be shared. When for instance a change in the business logic is required, either via configuration or via a new version of the software, these changes can be installed centrally and will be available instantly for all users. The main disadvantage is the susceptibility to peak usage. When all users request access to the same element at the same time, an overload may occur, and the mainframe may become the bottleneck in the organizational processes. Instances of peak traffic can occur during a financial month end, or when a successful marketing campaign results in an unexpected high number of orders.
The mainframe architecture has already existed for many decades. This does not mean that it is outdated. The principle of sharing disk space, internal memory and processing power is still valid today, and centralization of architectural elements in a one data centre location has regained popularity in the past few years. For ERP, a centralized architecture is very suitable, because the ERP database and the ERP business logic can easily be shared by all ERP users. he decentralized dumb terminal however have limited functionality for modern standards, and because of the availability of cheap network capacity and the efficient graphical options of modern input and output devices it is no longer necessary to limit network traffic. Dumb terminals and keyboards have therefore been replaced by more advanced monitors, printers, mice and scanners.