TCB system

The TCB system enables the management and the control of multiple Unix servers. The TCB system helps to execute updates on multiple number of servers at once, and enables the central maintenance of the configuration files of the servers. If the functions and the programs of the systems differ by only slightly, then the mass management of the systems becomes easy with the help of TCB. The differences between the systems are also manageable with a little effort. The best example for the efficiency of the system is that the close to 200 servers we manage are operated by only five administrators without stoppages.

The operation of the system

The system generates the expected structure of the system from the central pattern (that consists of binaries, data, and the configuration files), and subsequently estimates its actual status on the remote computer through a secure channel. Comparing the actual and the required structure, a list is generated of what should be changed on the remote system to get into the required condition.

If there is a difference, then the administrator must decide what to do. There may be several reasons for the differences. If a change happened on the local system (for example a program was updated, or the administrator has changed something in the configuration), then he/she can send the alterations to the server. If the change happened remotely (for example the local administrator of one of the servers modified something in the configuration files), then after its reconfiguration the modifications can be brought up to the etalon configuration. If, however, something has been changed in the remote system for an unknown reason, (for example one of the server programs has changed), that implies an attack, and the administrator must investigate what has happened.

The TCB system has the following properties (without trying to list them all):

• alteration of files in the etalon and the remote system at the same time,
• tracking the configuration changes (a comment has to be added in the case of every modification),
• mass execution of commands or checks on servers,
• meta-package management,
• regular checks, local penetration tests,
• can be conveniently fitted into the monitoring system,
• comparison of the files between the server and the etalon with one command.