Books
in black and white
Main menu
Share a book About us Home
Books
Biology Business Chemistry Computers Culture Economics Fiction Games Guide History Management Mathematical Medicine Mental Fitnes Physics Psychology Scince Sport Technics
Ads

Teradata RDBMS Database Administration - NCR

NCR Teradata RDBMS Database Administration - NCR , 2004. - 616 p.
Download (direct link): teradatadatabaseadmin2004.pdf
Previous << 1 .. 104 105 106 107 108 109 < 110 > 111 112 113 114 115 116 .. 218 >> Next


DSU/AMP 4

4, 12, 20 17, 18, 19

DSU/AMP 8

8, 16, 24 21, 22, 23

The following figure illustrates a configuration that consists of a single eight-AMP cluster configuration.

In this cluster, primary and fallback copies of table rows are distributed across all eight AMPs.

FF06A005

7 - 32

Teradata RDBMS Database Administration Chapter 7: Protecting Data

AMP Clustering and Fallback

DSU/AMP 1

DSU/AMP 2

DSU/AMP 3

DSU/AMP 4

Primary Copy Area Fallback Copy Area

DSU/AMP 5

DSU/AMP 6

DSU/AMP 7

DSU/AMP 8

Primary Copy Area Fallback Copy Area

FF06A004

The fallback copy of a row never resides on the same AMP as the primary copy. For example, if AMP 3 fails, its data remains available on AMPs 4, 5, and 6. However, if AMP 3 and AMP 5 both fail at the same time, row 11 is unavailable.

Note: Fallback-protected tables take up twice the space on your system as non-fallback-protected tables. For 7x24 systems, the fallback option on important tables is recommended for minimizing the risks of system downtime.

For information on the performance impact of fallback, see Teradata RDBMS Performance Optimization.

Failure Handling

If two physical disks in the same rank fail, the associated AMP fails.

When an AMP fails, the system reads all rows it needs from the disks in the remaining AMPs in the cluster. If it needs to find a primary row from the failed AMP, it reads the fallback copy of that row on a disk in another AMP. The system updates fallback copies of rows.

To repair the failed AMP, you must replace the failed physical disks, and reconstruct the data from fallback copies of rows on running AMPs in the cluster.

You can use the following utilities to reconstruct data:

7 - 32 Teradata RDBMS Database Administration

Chapter 7: Protecting Data

AMP Clustering and Fallback

Utility Description
RCVManager Allows you to monitor recovery processing
Table Rebuild Reconstructs tables on an AMP from data on other AMPs in the cluster.
Vproc Manager Allows you to display and modify vproc states, and initiate a Teradata RDBMS restart

See Teradata RDBMS Utilities for more information on these utilities.

Activating Fallback Protection

You can activate fallback protection via the CREATE/MODIFY DATABASE/ USER and CREATE/ALTER TABLE statements. Following is an example of a statement that activates fallback protection for a new user:

CREATE USER maxim ,AS PERMANENT = 1000000 ,PASSWORD = mxm ,FALLBACK;

Thus, fallback is the default for all tables created by Maxim in his own space. Maxim can override this default at the table level with the NO FALLBACK option of the CREATE TABLE or ALTER TABLE statement.

You can modify an existing database or user to use fallback:

MODIFY DATABASE AcctsRec AS FALLBACK;

However, MODIFY statements only affect tables that will be created in the future. Use ALTER TABLE to change the protection of existing tables.

Down-AMP Recovery Journal

The system uses the Down AMP Recovery Journal for fallback-protected rows when an AMP is out of service. An AMP is placed out of service if two physical disk fail in a single rank. The AMP remains out of service until you replace the disks and reconstruct the data via the Table Rebuild utility. See Teradata RDBMS Utilities for more information on the Table Rebuild utility.

Storage space for Down AMP Recovery Journal rows comes from user DBC PERM allocation.

This recovery journal keeps a copy of the tableIDs and rowIDs of fallback-protected rows on which changes have been made, and whose alternate copy (primary or fallback) would be under the control of the down AMP.

The Down AMP Recovery Journal is maintained by the other AMPs in the cluster. When the down AMP is returned to service, the system copies the indicated rows to disk from the other AMP, and discards the journal.

7 - 32

Teradata RDBMS Database Administration Chapter 7: Protecting Data

AMP Clustering and Fallback

Determining Cluster Size

In principle, the fewer AMPs in a cluster, the less likely that two AMPs in the cluster will fail at the same time. However, an AMP or its vdisk failure within a cluster dramatically increases the workload on operational AMPs in a cluster. Throughput is decreased and response time is increased. This situation is illustrated by the following graph.

2.0

U



1.0

2 4 6 8 12

Cluster Size in AMPs

16

Note that the failure of one AMP in a 2-AMP cluster increases the workload on the remaining AMP by 100%. However, a failed AMP in an 4-AMP cluster increases the workload on the remaining AMPs by only 33%.

Changing Cluster Assignments

You can use the Configuration utility to reassign clusters into larger groupings, depending on the size of your configuration and available disk space:

IF you ... THEN .
have a single-node system NCR recommends you retain the default cluster assignment.
change cluster assignments without adding AMPS or disks make certain ample disk space is available on all AMPs. A general formula for recommended space is: (CURRENTPERM * 3/2) should be less that 80% of total MAXPERM For guidelines, see "Adding Vprocs" on page 12-29. For detailed information on using the Configuration Utility, see Teradata RDBMS Utilities.
Previous << 1 .. 104 105 106 107 108 109 < 110 > 111 112 113 114 115 116 .. 218 >> Next