Fallback Cluster
A
cluster is a group of AMPs that act as a single Fallback unit. Clustering
has no effect on primary row distribution of the table, b ut the Fallback row
will always go to another AMP in the same cluster.
Cluster sizes are set through a Teradata console utility and may range
from 2 to 16 AMPs per cluster (not all clusters in a system have to be the
same size). The example shows an 8-AMP system set up in two clusters
of 4-AMPs each.
Should an AMP fail, the primary and Fallback row copies stored on that
AMP cannot be accessed. However, their alternate copies are available
through the other AMPs in the same cluster.
The loss of an AMP in one cluster has no effect upon other clusters. It is
possible to lose one AMP in each cluster and still have full access to all
Fallback-protected table data. If there are two AMP failures in the same
cluster, the entire Teradata system halts.
While an AMP is down, the remaining AMPs in the cluster must do their
own work plus the work of the down AMP. The larger the size of the
cluster, the less noticeable the workload increase is within that cluster
when one AMP fails. Large cluster sizes are more vulnerable to a second
failure before recovery from the first failure is complete. Remember that a
second failure halts the entire Teradata system.
If you had an 8-AMP system with 4 clusters of 2 AMPs each, the system
could lose four AMPs (one per cluster) and continue operations. If the
system workload is near capacity, there will be some loss of performance
No comments:
Post a Comment