Replication through the system journal is event-driven. When a data group is started, each process used in the replication path waits for its predetermined event to occur then begins its activity. The processes are interdependent and run concurrently. The system journal replication path in MIMIX uses the following processes:
-
Object send process: This process reads the journal entries added to the security audit journal. When an object identified in a journal entry is within the name space, the object send process creates a MIMIX construct called an activity entry, which is used to track replication status. The send process also determines whether any additional information is needed for replication, and transmits the activity entry to the target system. Data groups can be configured to use a shared object send job or to use a dedicated job.
-
Object receive process: This process receives the activity entry and waits for notification that any additional source system processing is complete before passing the activity entry to the object apply process.
-
Object retrieve process: If any additional information is needed for replication, the object retrieve process obtains it and places it in a container within a holding area. This process is also used when additional processing is required on the source system prior to transmission to the target system. The object retrieve process uses multiple asynchronous jobs. The minimum and maximum number of jobs is configurable for a data group.
-
Container send process: When any needed additional information has been retrieved, the container send process transmits it from a holding area to the target system. This process also updates the activity entry and notifies the object send process that the additional information is on the target system and the activity entry is ready to be applied. The container send and receive processes use multiple asynchronous jobs. The minimum and maximum number of jobs is configurable for a data group.
-
Container receive process: This process receives any needed additional information, places it into a holding area on the target system, and notifies the container send process when it completes these operations.
-
Object apply process: This process uses the information in the activity entry as well as any additional information that was transmitted to the target system to replicate the operation represented by the entry. The object apply process uses multiple asynchronous jobs. The minimum and maximum number of jobs is configurable for a data group.
-
Status send process: This process notifies the source system of the status of the replication.
-
Status receive process: This process updates the status on the source system and, if necessary, passes control information back to the object send process.
Each replication process removes unnecessary messages from its job log on a daily basis.
MIMIX uses a collection of structures and customized functions for controlling these structures during replication. Collectively the customized functions and structures are referred to as the work log. The structures in the work log consist of log spaces, work lists (implemented as user queues), and distribution status file.