Backup and recovery of DB2 Analytics Accelerator data now possible!

Overview

The IBM DB2 Analytics Accelerator for z/OS provides the ability to gain greater insights and analytics capabilities from their enterprise DB2 data.   The Accelerator allows you to run analytic queries against their existing DB2 data at higher speeds and lower costs than the same queries running natively on IBM DB2 for z/OS. This is great for data that exists today in DB2 databases, but many Accelerator customers are looking to exploit the Accelerator’s capabilities for more than DB2 data, such as VSAM, Oracle, DB2 LUW, SMF or many other types of data.  This can require a complicated and costly conversion process:

  • Convert the data to a format understood by DB2

  • Load the data to a DB2 for z/OS table

  • Load the data to the Accelerator  

This multi-step process can be cumbersome, and requires CPU and disk resources on z/OS since the data is operationally maintained and protected in its originating location.  There are also Accelerator customers today looking to build an operational data warehouse on DB2 for z/OS to take advantage of an attached Accelerator. 

 

Accelerator-only tables

As Accelerator usage has increased, support to handle more data types and queries has also risen. Some customers are asking the question “Do I need to load the data in both DB2 and the Accelerator if all my queries are analytical queries that can be run on the Accelerator?”  Loading data only to the Accelerator is possible with a new type of table called the Accelerator-only table (AOT).  This table is defined in and access is controlled by DB2 for z/OS, but the data resides only in the Accelerator.  All queries are automatically routed from DB2 for z/OS to the Accelerator.  Loading data to an AOT can be accomplished via a programmatic insert or by selecting data that already resides in the Accelerator and sending the result set to an AOT.  While this method works well for DB2 data, it can be challenging to load non-DB2 data directly to an AOT via insert statements. 

 

Background

The IBM DB2 Analytics Accelerator Loader for z/OS, (Loader), provides an easy way for Accelerator customers to extend their analytic insights from data that originates or is maintained in many different formats and locations.  This includes VSAM, IMS, Oracle, DB2 LUW, z/OS operational data such as SMF and many other types of enterprise data.  The Loader provides a cost-effective way to reduce the multi-step process to a single step.  Data can be directly loaded to AOTs or to a standard DB2 and/or Accelerator-shadow table in parallel.  This provides greater flexibility over the type of table that is used.  If all of the query workloads can be run by the Accelerator, an AOT makes perfect sense.   A major concern has been a lack of a backup and recovery capability when data is loaded to an AOT.  Though the Loader can perform a ‘Load Resume’ (adding data to an Accelerator table), there is a significant investment over time in this loaded data that would be costly to recreate.

 

You may also want to retain the data in the Accelerator longer than it is retained at the original source.   With high compression rates and large storage capacity, the Accelerator is a good solution for storing and performing analytics against historical data in real-time.   Some customers implemented AOTs with the knowledge that the data is operationally maintained at its origin, but quickly found out the investment they made in loading this data over time is at risk without a good backup and recovery strategy.  

 

The Solution

In response to customer requests, the IBM DB2 Analytics Accelerator Loader now includes full backup and recovery functionality for Accelerator tables including AOTs (Available via APAR PI70981/PTF UI44867).  This new function includes the ability to create an INLINE copy while loading data, and a standalone backup utility to create a backup copy of the data.  The copy is stored on z/OS and can be written to a virtual tape system.  The Loader keeps track of all the copies in a ‘SYSCOPY-like’ table (HLOUCOPY)and provides a recovery function to restore the data by selecting one or more copies.  The design of the backup and recovery function was meant to be very DB2 for z/OS DBA friendly.  Many of the concepts of a DB2 for z/OS backup and recovery apply.   All backup and recovery functions are invoked via DB2 for z/OS; no direct access to an Accelerator is required. Both full and incremental copies and INLINE copies are supported, and up to four copies can be created, two for local and two for remote recovery. 

Views: 392

Comment

You need to be a member of The World of DB2 to add comments!

Join The World of DB2

© 2017   Created by Surekha Parekh.   Powered by

Badges  |  Report an Issue  |  Terms of Service