Archives of the TeradataForum

Message Posted: Mon, 07 Nov 2005 @ 18:11:00 GMT

	<Prev	Next>		<<First	<Prev	Next>	Last>>

In response, to your question, cluster restores can (and should) be parallelized. If you have 100 vprocs (amps) you may have as many as 50 clusters if you have only two amps per cluster. You could start 5 ESCON connections and backup 10 clusters per connection. You have to ensure that each output file is a different name per connection. That should give your backup a throughput of (5 x 30 Gb) or 150 Gb per hour...

It will depend upon how many ESCON connections you wish to establish then divide that number into the number of clusters configured on your system, that will tell you many clusters to back up on one data stream. Also, in changing to a cluster backup, you need to remember to backup the DBC tables as a single stream separately, as a cluster backup will exclude the DBC automatically. It is also good insurance to add a Dictionary only Backup stream as well.

The restore would consist of 1) Restore the DBC, 2) restore the Dictionary, 3) parallel restore of the cluster data, 4) Execute the BUILD command for all tables restored.

The degree of parallelization achieving high throughput is greatest with very large tables that have little or no skew!

Best Regards,

Michael E. McBride, MSCIS

	<Prev	Next>		<<First	<Prev	Next>	Last>>