we have bar solution which runs arcmain behind the scene and even on 4 TB it takes more than 20 hours how can we minimize this 20 hours and some times before the completion of backup new load jobs started ? How can we implement incremental backup (a complete guidance link for that ?)
Not really enough information to give an intelligent response, other than to suggest looking into Teradata Advocated BAR solutions.
nb. ARC based backup solutions to not support any real form of incremental backup
DSA is the replacement technology which does support a form of incremental backup ie. Change Block Backups (CBB), note that like many other backup functions that dramatically reduce the backup window the cost comes at the restore ie. all restores of the complete object to ensure its integrity.
Backup of 4TB in 20 hours is 200MB/hour, a terible poor speed even for single stream.
Many things can be the bottleneck here;
The Teradata system; is it a SMP, a MPP, a VM, on AWS ?
How busy is the system during backup; idle, moderate, saturated ?
The tape system; is it advocated BAR, is it self provided, is it to disk, to tape ?
The archive job; is it single stream, multi-stream, database level, table level ?
What speed do you achieve with a single table in a single job ?
Many thing to look into. Without any more detail, I would suggest to split the job into multiple jobs and run multiple backup streams in parallel.