By Philip E. Courtney
Recovery of distributed data is one of the
most pressing concerns for contingency planners. While the stand-alone
mainframe environment has long possessed mature tools designed to centralize
backup and recovery of application data, the distribution of data on PCs
and LANs scattered throughout the corporate computing enterprise has served
to dramatically complicate the recovery process. In many companies, the
number of procedures employed for backup and recovery are equal to the
number of LANs and LAN administrators. Yet, despite the multitude of procedures,
few companies can safely admit to employing comprehensive recovery capabilities.
Possessing an extensive network of mainframe-connected PCs and LANs using
a combination of IBM’s OS/2 LAN servers and Novell Netware, Farmland Industries
maintains the comprehensive capability to recover application and corporate
data stored on any PC and LAN server. Employing FDR/UPSTREAM (Innovation
Data Processing, Little Falls, NJ), Farmland has dramatically simplified
the backup process for distributed data by implementing automated, unattended
and centralized recovery management. "FDR/UPSTREAM facilitates the backup
and recovery of data stored on the LAN servers," says Dan Neisen, communications
analyst programmer for the Kansas City, MO-based agricultural company.
"The product tracks data set name, file and data block count as well as
storage means. It also provides a secure method for managing the recovery
of distributed data."
Prior to implementing centralized recovery
management, Farmland was faced with the problem of increased data and
limited storage capabilities for recovery. According to Neisen, the tape
backup process formerly employed for capturing PC and LAN server data
was cumbersome, ineffective and often unreliable. "Our previous tape backup
method was unable to process in a timely fashion the large volume of data
stored on our LANs," he explains. "In many cases, we were uncertain if
the backup had occurred properly. It was a guessing game as to whether
we could get files back."
During the past several years, many companies have taken advantage of
declining PC hardware costs by installing machines with greater processing
power and dramatically increased storage capacities. Compounding the data
integrity challenge for recovery analysts was the expanding number of
networks scattered throughout every corporate division, each containing
many thousands of megabytes of data. Combined with improved communications
and connectivity technology, an ever-increasing amount of vital corporate
data was stored in a distributed environment sometimes rivaling - and
often exceeding - the amount of data stored on the centralized mainframe.
While tape backup systems exist for individual PCs or LAN servers, these
systems are often unable to appropriately synchronize backups and recoveries
across multiple PCs or LANs - an especially critical function when analysts
strive to achieve a time-based recovery. Synchronized backups are equally
critical in environments performing PC-based development for mainframe
applications to ensure an accurate recovery. Furthermore, diskette backups
of individual PC data are unreliable and manual, consistently requiring
the user to insert new diskettes while also possessing an inherent security
"The multitude of backup procedures presented a tremendous exposure risk,"
says Neisen. "We required a method to ensure the proper recovery of all
data on PCs and LANs and combine those recovery steps with our existing
mainframe recovery procedures. Centralizing management was a service we
could provide to our corporate personnel."
The implementation of FDR/UPSTREAM helped Farmland centralize recovery
management for PC- and LAN- based data. Farmland has implemented both
MVS and distributed components linked via Advanced Program-to-Program
Communications (APPC) LU6.2 to capture the data distributed across the
entire corporate computing enterprise for storage on the mainframe. According
to Neisen, data copied to the mainframe is compressed by the product for
rapid transfer and simply incorporated into existing disaster recovery
procedures. "The files are placed on either DASD or tape and managed just
like a mainframe file," he explains. "The product supports a multitude
of storage options."
While the stand-alone mainframe environment has long possessed mature
tools designed to centralize backup and recovery, the distribution of
data on aPCs and LANs scattered throughout the enterprise has served to
dramatically complicate the recovery process.
These options, Neisen contends, enable Farmland to adjust storage mechanisms
based on application requirements. For example, while some distributed
data can be placed into a dynamically allocated mainframe sequential file,
other forms of on-line DASD storage include a single repository managed
by FDR/UPSTREAM or multiple systems-managed storage-controlled files.
At the discretion of each application or site, data in these files can
be automatically "batched" and transferred to tape at a later date to
either free DASD space or send information to an off-site storage location.
Application-critical data can also be written directly to a dynamically
allocated mainframe tape, which can be under the control of an existing
mainframe tape management system. "Our tape management system controls
the tapes once the data is transferred to the mainframe," says Neisen.
Expiration dates for tapes are established via GDG cycles and are scratched
by CA-TLMS (Computer Associates International, Islandia, NY) when cycling
Communication between the PC and mainframe components ensures that only
the appropriate files are transmitted to the MVS component. The product
also contains version control logic that maintains information regarding
the status and location of all data. This logic provides all corporate
personnel with the ability to perform an authorized on-line inquiry of
individual backup parameters and file-set specifications to retrieve such
information as archive status, backup completion, Novell (Provo, UT) and
BANYAN (Banyan Corp., Westborough, MA) data on any individual file or
group of files. Additionally, file inquiries enable the display of all
directory files including file name, size, backup date and so on.
The Farmland distributed recovery procedures include daily incremental
backups and semiweekly merge backups. For the incremental backups, the
data is stored for a single generation on DASD while older versions are
copied to tape by CA-TLMS. "Sufficient overlap exists so we can rapidly
recover and rebuild our files in the event of a disaster," says Neisen.
This overlap also ensures corporate personnel can immediately retrieve
and recover different versions of data from the preceding several days
as input to other applications. Equally important is the ability to further
mirror mainframe archiving capabilities where data remaining unused for
a specified time period can be automatically copied to mainframe storage
and removed from the PC or LAN.
According to Neisen, FDR/UPSTREAM has integrated
with the company’s existing environment, providing the ability for completely
unattended and automated distributed data backups. Working in conjunction
with existing tools, Farmland’s backup procedures are initiated from the
mainframe by the Job Trac (Legent Corp., Herndon, VA) job scheduling system.
"All backups are host-initiated," explains Neisen. "Job Trac submits the
appropriate jobs for FDR/UPSTREAM to process. We also have the capability
to submit jobs from the server and operate in either attended or unattended
By employing both the mainframe and server components of the product,
Farmland has improved recovery capabilities by customizing the backup
parameters to meet the needs of each individual application. While the
mainframe component contains the control parameters indicating the Logical
Unit (LU) names of the PCs to be processed, the server component contains
the actual backup parameters - incremental, archive, merge, full volume
- for individual and groups of files. "FDR/UPSTREAM from the PC determines
what should be backed-up while the component on the mainframe determines
how that information should be handled," says Neisen.
Employing an LU6.2 conversation, the jobs submitted by Job Trac communicate
with the mainframe FDR/UPSTREAM started task. The started task, which
communicates via APPC to the LAN components, operates as an "on-line initiator"
and causes the actual backup process to begin at the server or workstation
level. "The process is automated and unattended," says Neisen. "Accurate
and secure backups can occur at any time of the day without requiring
the presence of any personnel."
The files processed by the Farmland procedures represent a cross section
of the files present in virtually every organization. From application
development and end-user decision support databases to files created from
a variety of application using IBM’s OS/2, Microsoft Windows, WordPerfect
and DrawPerfect (WordPerfect Corp., Orem UT), Farmland has ensured the
successful recovery of all corporate data. "It doesn’t matter much what
files the corporation employs in the distributed environment," says Neisen.
"We can get them all."
The procedures implemented by Farmland have simplified the recovery of
distributed data. Using the inquiry capabilities contained within the
product, corporate personnel can merely select the individual file or
groups of files necessary for a restore. Furthermore, complete server
restores are possible in the event of a major catastrophe. "The restores
performed at our host site are typically for lost or corrupted files,"
Neisen explains. The company also performs full volume restores for locally
attached servers in the event of disk crashes. "We have also successfully
performed full server restores while testing at our disaster recovery
hot site," he says.
Neisen notes that another advantage of the restore capabilities contained
within the product is that corporate business units possess the ability
to share data even on independent LANs. With the restore function, corporate
personnel now possess the ability to retrieve necessary information from
the mainframe or server "warehouse." "The data is also much more secure,"
Neisen says. "Our mainframe security controls access to the files stored
on tape or DASD, while LAN security provides the necessary mechanisms
to manage access in the distributed environment."
Centralizing the management of distributed data recovery has eliminated
many of the redundant tasks performed by Farmland LAN administrators.
According to Neisen, FDR/UPSTREAM has simplified administrative tasks,
enabling technical analysts to specify the necessary LU names and necessary
backup criteria during product implementation. "Centralized administration
assures us the backups are always performed," he says.
Most important, he says, is that Farmland now maintains the comprehensive
capability of recovering all of the vital corporate data distributed throughout
the entire enterprise. "With all PCs connected to the mainframe," says
Neisen, "not a single machine can elude the backup process."
ABOUT THE AUTHOR
Philip E. Courtney is a freelance writer based in Bushkill, PA.