Thursday 6 March 2014

Fast load

FASTLOAD
  • ·         Fast load is a batch mode utility to load huge volume of data as fast as possible
  • ·         Compare with all utilities Fast load is faster
  • ·         Fast load loads errors into Error Tables
  • ·         It is fully Automatic Restart ability and checkpoint configurable
  • ·         Fast load not  support duplicate record loading if table is multi set and table should be empty.
  • ·         It runs in 2 phases and process the data block by block.

·         Fast load require Two Error tables to catch any error which occurs during the execution of the Fast load.
Error Table1:
Error Table1 contains
1.   Constraint violation errors
2.   Unique violation errors
3.   Unavailable AMP errors
Error Table2:
Error Table2 contains the errors if the table has UPI on the Table and we are trying to load duplicate records this type of errors loads into error table2.
Log Table:
  • ·         Fast load internally creates log table to implement the restart ability capability.
  • ·         If fast load script failed it stored failed point in log table.
  • ·         When script restarts it runs from the last failed point stored in log table.


Fast Load runs in Two Phases

Phase1:
·         The main objective of this phase is to send rows of data file from HOST COMPUTER to Teradata AMPs as fast as possible.
·         Rows of the data file packed into 64k Blocks and send it to PE.   
·         PE parse the SQL of Fast load and send the explain plan to each AMP. It will by default create 1 session for each AMP, so if u r system has 100 AMP then it will make 100 sessions for fast load job.
·         After creation of sessions the 64blocks of data is passed to AMP with the help of PE and BYNET where it is quickly hashed according to its PI value.
·         Based on the row hash value the rows are then redistributed its proper AMP.
·         Now each row is placed in this proper AMP, but they are not sorted still now.
·         Errors in this phase load Error Table1.

Phase2:
·         The main objective of this phase to store each row into the Actual Target Table
·         The rows sorted by AMP which is temporarily stored in its DISK during phase1
·         These sorted then send to actual target Table where they will reside permanently
·         Any errors in this phase stored into error table2



          

No comments:

Post a Comment