Architecture of Data Pump

processes are

  Process in Data Pump

Master Process

• The mcp has a process of format Instance_DMnn_pid
• For Every Data pump export or import job a master process is created.

• It is responsible for communicating with clients,
• Creating and controlling a pool of worker process
• And performing logging operation.
  Creates jobs and controls them
  Creates and manages worker process
  Monitor jobs and log progress
  Creates the master table in the user schema
  Manages the necessary files including the dump file set

The worker process

• Worker process is responsible for heavy duty workload of loading and unloading data and has the name of Instance_dwnn_pid .
• The mcp creates the worker process . The number of worker process depends on degree of parallelism of the job.
• The woker process maintain the object rows of the master table.
• It updates the master table about information about status of various jobs, completed ,pending or failed.

  Shadow process

• When a client logs into Oracle server the database creates an Oracle foreground process to service the Data Pump Api requests. The shadow process creates the job consisting of master table as well as master process
• Once the client detaches the shadow process automatically disappears
•  
• The shadow process creates a job which primarily consists of creating the Master Table creating the queues in Advance queue used for communication among various process and creating the master control process.
•  
• Once a job is running the shadow process main job is to check the job status for the client process.
• Client process

• Client process (expdp or impdp ) calls data pump api .
• After data pump job is established the client process is not required to keep the job running.
•  

Parallel Query Process

This process is used when Data Pump Chooses External Table Api as the data access method for loading and unloading data

.

Data pump Loading Detail

• The open call starts the MCP Process and creates two AQ queues.
• The first queue is status queue used to send status of jobs. Which includes logging information and errors.
•  
• Clients interested in status of the job can query this queue.
• This is a unidirectional queue.
•   MCP Posts the information to the queue. And clients consume the information.
•   The second queue is command and control queue which is used to control the worker process established by mcp and to perform api and file management.
•  
• MCP Directs one work per process to do meta data upload
• Using dbms_metadata api
• Master table is written to dump file at end of every export and it is loaded from dumpfile to user schema at import and it used to find the sequence of tables to be loaded during import.

Advantages of Data Pump

• Data Pump can be restarted
• Server Side Parallel Execution
• Automatic Performance Tuning
• No need to send buffer size – etc to improve performance as you do in Export and Import
• Simplified Monitoring
• Object Filtering
• Can filter full range of object types
• Can remap tablespace and datafiles on import
• Estimate the export file size (Either using statistics on Block size)
• Paralleliszable
• Significanty Faster than tranditional export /import
• A file is not required while importing through network link
• Monitor how much time required and how much is spent etc
• On v$session_longups
• Resumbale interruptible and restart able
• Status of the Job
• Status shows how much is the progress of job.
• Transport_Full_Check
• Transport_Tablespace

o email info@cpisolution.com Contact NumCallr:0703 180 5732