Most large companies have a mix of multiplatform IT environments: mainframe, Unix, Linux and Windows. Typically, these environments don't share software, staff, and disk and tape resources. But with a few caveats, there are considerable advantages to using a mainframe to back up all of an organization's data to streamline backup and disaster recovery (DR) operations.
Take critical applications, for example, that mostly run on integrated multiplatform systems. An enterprise resource planning (ERP) application such as SAP includes back-end databases running on the mainframe, application processing on Unix servers and Web interfaces running on Windows. There's usually a large amount of data transferred between these environments via FTP, extracts and dumps. This creates a lot of overhead and, at times, impacts the availability and service-level agreements of the business applications. In addition, DR plans for these critical applications are sometimes separate for each storage environment. This makes the data more difficult or impossible to recover because of the following:
- Parts of the data are backed up at different times and aren't synchronized.
- Data is backed up using different tools such as Innovation Data Processing's (IDP) FDR on the mainframe, IBM Tivoli's Storage Manager on Unix and Veritas Software Corp.'s Backup Exec for Windows Servers.
- The data is placed on different tape drives and media such as IBM's 3590 tape drives for the mainframe, Storage Technology Corp.'s (StorageTek) 9940 on Unix and LTO tapes on Windows. There are also different tape handling processes, labeling, vaulting and staff.
A consistent view of data
If the mainframe is used to increase recovery success rates for most mixed environments, there must be a consistent view of the data at a point in time, called consistency groups. The ability to gather data from various platforms at the same time is available with products from most major mainframe/open storage vendors, such as EMC Corp., Hitachi Data Systems (HDS) Inc., IBM Corp. and StorageTek. EMC's TimeFinder and HDS' ShadowImage can coordinate I/O freezing for a database at the host level. IBM's FlashCopy performs this task at the array level. Achieving tight coordination of hosts, applications and data access requires scripting and programming job-scheduling packages to make sure the application is quiesced and data is flushed.
For example, EMC's solution for Symmetrix storage environments combines its TimeFinder software with IDP's FDRSOS and FDR/Upstream/SOS. The combination provides a point-in-time copy of critical data across multiple platforms.
The diagram "Mainframe backs up all the data" shows Unix and Windows systems connected to an EMC Symmetrix array. The volumes are set up with business-continuance volumes (BCVs) using TimeFinder. The BCV data is visible to the mainframe and is backed up using FDRSOS, which is a z/OS- or OS/390-based solution for physical volume, high-speed backup/restore of open systems resident on EMC storage over existing mainframe ESCON or FICON channels. Most users will also use the FDR/Upstream/SOS file-level component for more granular backups and restores.
FDR/Upstream/SOS reads the data off the open-systems disk (which can be a storage area network [SAN] disk, local disk or a LAN logical disk) and writes it to the specially formatted FDRSOS disk on the SAN. The FDR/Upstream OS/390 component then takes the data off the SAN disk and writes it to tape. This process is transparent to the administrator and results in a highly reliable, high-speed solution for LAN and SAN environments. Backups don't go across the network, but travel through the mainframe's ESCON or FICON connection. And open-systems storage administrators aren't held hostage by a mainframe-only process because they can perform their own backups or file-level recovery using a Java-based browser interface.
Organizations can orchestrate all of the dependent steps using mainframe job schedulers such as Allen Systems Group (ASG) Inc.'s ASG-Zeke, Computer Associates (CA) International Inc.'s Unicenter CA-Jobtrac Job Management, IBM's CA7 utility and open-systems schedulers such as CRON. With this approach, data is backed up using mainframe resources and staff in a consistent manner. This approach also provides the benefits of LAN- and server-free backups. All the work of backup and recovery is done by the mainframe, so fewer open-systems resources are used.
John Hunter, senior systems programmer at Sigma-Aldrich Corp., St. Louis, MO, uses the mainframe to back up data. "By using FDR/Upstream/SOS, we were able to consolidate our entire Oracle, Windows, Unix, Novell and Domino backups," he says. "Administrators for each platform install a client, and our mainframe manages scheduling and tape management, giving us a backup solution with a single point of control."
However, there are potential downsides to handing off open-systems backups to big iron and tying all the enterprise's back up to a centralized, high-cost solution. For the mainframe to successfully back up open-systems-based applications, there needs to be a high level of coordination and communication between staffers from both camps. As a result of isolated responsibilities and differences in the way the two groups view storage, in many companies the communication between mainframe and open-systems personnel isn't very good. It also takes a great deal of time to set up a mainframe backup process for the entire enterprise storage environment, although some would argue that it is simpler to set up and run than having, for example, three backup packages, three types of media and more than three different procedures. And, of course, mainframe backup costs more than open-systems backup; however, depending on how total cost of ownership is calculated, mainframe backup could save money if critical data and apps are more quickly, and thoroughly, restored after a disaster.
