Mainframe Appliance for Storage
White Paper

Bus-Tech's Mainframe Appliance for Storage

Mainframe Appliance for Storage is a virtual tape controller for IBM System/390 and compatible mainframes.

This paper introduces the reader to the concepts of virtual tape as well as the solution offered by Bus-Tech and then reviews several of the applications where Mainframe Appliance for Storage can provide immediate benefit.


Virtual Tape Overview

Virtual tape systems (VTS) are often referred to as "tape-on-disk" because they emulate physical tape drives and volumes using permanently mounted disk storage. Data being written to a tape volume in a VTS is actually stored as one of many files on a disk. A defining characteristic of virtual tape systems is that they are transparent to the host computer and application software they are attached to. In other words a VTS must appear to the host computer as one or more real tape drives; responding to all processing requests exactly as a tape drive would. The host operating environment and any application software using the VTS cannot be aware that the data is being stored on disk rather than tape.

Virtual tape systems typically achieve return on investment by:

  • Reducing / eliminating the consumption of expensive tape cartridges
  • Shortening batch processing windows by eliminating tape mount requests and increasing performance
  • Reducing/eliminating high, re-occurring maintenance charges on tape drives
  • Reducing physical space allocated to tape storage and tape drives
  • Reducing personnel costs associated with managing a large tape library

And because Virtual Tape Systems are transparent to the host computer, benefits can be realized without re-engineering and deploying business application's software.


Mainframe Appliance for Storage

The Mainframe Appliance for Storage (MAS) is a virtual tape controller connecting System/390 class mainframes to open systems storage. MAS uses one or two Enterprise System Connections (ESCON) to connect to System/390. Each ESCON can emulate 1 to 16 IBM 3480 tapes drives. Although the MAS uses 3480 emulation as its basis, it is not restricted to the speeds or cartridge size limitations of real 3480s. Virtual cartridge size is defaulted at 2 GB, but can be overridden by a user parameter. Jobs/tasks running under z/OS, OS/390, or VSE can allocate tape volumes on emulated tape drives in the same fashion they would allocate a real tape drive. Tape management systems operate transparently with MAS to control tape volume allocation and expiration just as with real tape volumes.

The MAS stores tape data received from the mainframe on externally attached direct access storage (DAS) in a folder (directory) known as the virtual tape library. External storage can be attached to MAS via ultra SCSI or fibre channel connections and can range from a single disk to RAID (redundant array of independent disks) to a partition or segment of a larger disk subsystem such as IBM's Enterprise Storage Server (Shark) or EMC's Symmetrix.

The MAS is a 2U (3.5inch) high rack mount controller designed to install into an industry standard 19 inch rack. The MAS comes standard with a single ESCON interface and an ultraSCSI port. The MAS has internal, mirrored disks which house the MAS embedded Linux operating system and virtual tape software. The MAS requires a monitor, keyboard, and mouse which are used as the MAS operator console. The MAS operator console is used to:

  • Configure the MAS Virtual Tape Drives
  • Configure the attached direct access storage (disk)
  • Monitor the operation of the MAS
  • Provide maintenance on the virtual tape library.

The MAS can be enhanced with a second ESCON interface and/or fibre channel connectors.

How Does MAS Work?

To the mainframe each ESCON interface on the MAS appears like 1 - 16 3480 tape drives. MAS provides it's own Unit Information Module (UIM) for the Hardware Configuration Definition (HCD) utility allowing virtual tape drives to be collectively named "VTAPE" in job control statements in z/OS or OS/390. Once the MAS is defined to the system a job simply has to indicate "UNIT=VTAPE" on a data set definition statement and the MAS will allocate a virtual tape drive to the job.

The MAS uses unique leading identifiers in the volume serial number to identify the tape files that belong to it. A one or two digit prefix in the vol-ser field identifies these files. The user might assign a range of Vxxxxx to identify virtual volumes that are assigned to the MAS. Any number of unique identifiers can be used to accommodate a higher number of virtual cartridges or to segregate vol-sers by application (i.e. Pxxxxx for Payroll). The MAS uses these unique ranges of tape volume serial numbers to allocate new tape volumes. In z/OS and OS/390 the MAS monitors devices for a "load mount" request from the mainframe. When a scratch tape is requested to create a new volume, the MAS simply looks for the first available volume serial number within range, opens a disk file using the serial number as the file name, and then responds to the operating system's mount request.

When MAS receives a load mount request with a specific volume serial number, it checks to see if there is a file on disk with the serial number requested and mounts it.

In VSE, where the operating system does not send load mount requests, Bus-Tech provides utilities which can be invoked by any job to issue a load mount. Bus-Tech utilities include support for both Computer Associates' Dynam/T and Bi Moyle Associates' BIM-EPIC tape management systems; allowing jobs to request a data set by filename without knowing the volume serial number. The latest release of BIM-EPIC software offers native support for the MAS.

Maintaining Data as Written

System/390 tape volumes contain several different types of records including tape marks separating different files on the volume, header records, trailer records, and data records. Each of these records are different lengths and formats and a single tape volume may contain multiple data files; each with headers, trailers, and data. Within a single data file each data record may be of different length.

Virtual tape systems, like the MAS, must be capable of reproducing a tape volume during a read operation from the mainframe exactly as it was written in the previous write operation.

Re-creation of previously written tape volumes is achieved by MAS using an IBM standard file format known as "AWSTAPE". Each AWSTAPE disk file emulates one physical tape volume and contains an exact replica of data written to the tape including tapemarks, headers, trailers, and data files.

Each physical record written to the tape volume is represented on disk by a block header followed by the data record. A tapemark is represented by a block header only. The block header includes the length of the data record to follow; allowing MAS to recreate the tape record exactly as it was written.

Data encoding within a data record is left unchanged by MAS. That is, data records written by the mainframe in EBCDIC are written to disk in EBCDIC without modification. Data records written by the mainframe in ASCII are written to disk in ASCII.

Maintaining the Tape Library

Most installations rely on mainframe-based tape management software to control their tape library. Tapes (cartridges) are formatted with a volume serial number written in the first record of the tape. The volume serial number is used by the tape management system to keep track of what is physically written to the tape and when or if it may be re-used.

So, for example, it would be the tape management system that would keep track of which cartridge (by serial number) contained this week's payroll backup along with other information such as how many backups must be maintained and when a backup may be scratched.

Periodically the tape management system produces a scratch report indicating which cartridges (again by serial number) are available for re-use. The computer operations staff then pulls the cartridge from the tape library and sends it to the computer room where it can be mounted in response to a scratch request. The tape management software, seeing the volume mounted, knows that it is a scratch volume and allows data to be written over whatever was previously there. The tape management database is updated to indicate the new file (or files) on the volume.

Virtual tape volumes that are scratched by the mainframe's tape management system must be deleted from the MAS virtual tape library in order to free disk space and make the volume serial number available for re-use. This is basically equivalent to an operator pulling the tape from the tape library. Deletion of "scratched" virtual tapes can be accomplished in one of two ways:

  1. An operator may use a MAS directory utility and manually flag volumes for deletion form the MAS operator's console.
  2. A mainframe-based delete utility supplied with the MAS can be run. The utility will read a tape management scratch report, identify virtual volumes, and send delete requests to the MAS.

MAS Application

Now that you have been given an overview of how MAS works, it is time to take a look at how MAS may be applied within the mainframe environment and what potential benefits it might provide.

Enhancing Daily Backup

Most mainframe installations today have demanding online processing windows where online systems must be available to networked users. This demand for online systems creates fixed batch processing windows, usually during the night, when all offline activities must be completed.

For many installations, planning of batch processing is dictated by the need to backup large online databases within a fixed processing window. Often configuration of a system's tape hardware is determined by this underlying need.

MAS can enhance the daily backup process by moving daily backups from real tape to virtural tape; separating backup of the online databases from the process of writing tape/cartridges. As shown above a fibre channel equipped MAS uses an industry standard fibre channel disk array to house the virtual tape library.

During the daily backup of mainframe databases, virtual tapes are built in the virtual tape library on the array. Mainframe tape drives are free for other purposes. Backup processing times are shortened by two factors.

Tape cartridges only hold so much data. Large database backups typically require multiple volumes be mounted. Virtual tape volumes support larger files reducing the number of volumes required and the processing demand on the mainframe. Additionally, since the volumes are now virtual, mount request time is reduced to seconds. These two factors have shown to shortened backup processing windows by as much as 35%.

Once the mainframe backup process is complete online resources can be placed back online. If removable backup is required for offsite storage, the backup of virtual tape volumes can be accomplished using MAS-based backup utilities with no impact to online mainframe applications. A MAS dump utility allows individual virtual tape volumes or the entire tape library to be copied from the disk array to a SAN or SCSI-attached tape subsystem. Such MAS-based backups can be set up to occur automatically at a specific time during the day and make efficient use of tape cartridges by writing multiple virtual tapes to a single cartridge.

Alternatively, if a SAN-based backup server is being used, backup of the virtual tape library can be routinely scheduled as part of a larger SAN-based backup.

Finally, if a real 34xx tape cartridge is required for transfer to another mainframe, simple mainframe-based utilities can be used to read one or more virtual tape volumes and copy them to a real cartridge.

Batch Processing

Historically, the high cost of mainframe direct access storage (DASD) dictated that large batch processing applications were designed using tape. Processing involved reading one set of data from tape, performing necessary processing, and then writing a new version of the data set to a new tape volume.

Over time the demand for online database access and reduced costs of DASD have moved applications away from the tape batch processing model, but not entirely. The high labor cost of re-engineering major business applications has meant that some batch applications have been left on tape and their online databases are built in a separate processing step once updated tape volumes have been created.

By reducing the number of volumes required to hold datasets and eliminating manual operator mount requests, MAS shortens the processing windows of tape-based batch applications. More importantly, these shorter processing windows can be realized without a major re-engineering effort to re-write the applications.

Data Exchange

Most mainframe installations today have mixed environments including Unix, Linux, and Windows servers in addition to their System/390s.

Often there is a need to periodically move bulk data between the System/390 and one of these other environments.

A SAN-attached MAS can be used to periodically move data from System/390 to another server.

By default all virtual tape drives on the MAS write and read from the same virtual tape library (a file folder named "tapelib"). However, the MAS Configuration application allows a virtual tape drive to write data outside of the virtual library.

Using a file system common to both the MAS and the open server (Microsoft's FAT32, for example), it is possible to temporarily mount a SAN-based disk on MAS and use it to transfer a file to another server. Care must be taken when doing this since fibre channel SANs do not provide mechanisms to prevent data corruption if two servers write to the same disk at the same time. But as long as manual controls are used, sending data to an open server using the MAS is straight forward and easy. MAS can replace expensive mainframe and server-based networking applications that provide data exchange through the corporate network.

z/OS and OS/390 can automatically convert data from EBCDIC to ASCII as it is written to the virtual tape. Once the file has been created on a SAN disk, the receiving server can use any application or utility capable of supporting AWSTAPE format to retrieve the data.

Data Exchange in a NAS environment, utilizing a gigabit Ethernet connection and supporting a flat file format, is even more straightforward.

Flat file support will allow a virtual tape drive to output a file as a standard "flat" file rather than in AWSTAPE format. Targeted specifically at data exchange, flat files are not easily re-read by the mainframe but can be processed by other servers using standard utilities like Unix/Linux dd.

The mainframe data is written out as a standard tape operation to the MAS. The MAS, by the use of Samba, is defined as an NFS client and the data is sent to the targeted UNIX platform (NFS mount point) using standard file transfer protocols. In a Windows environment, CIFS is supported in the same manner as NFS.

The figure below demonstrates the file exchange capabilities of the MAS with gigabit Ethernet connections.

Gigabit networking can be added to any MAS and will provide support for CIFS (Windows file sharing), NFS (Unix/Linux networked file system), and iSCSI protocols. This will allow MAS to use a networked disk for both the virtual library and / or for data exchange. Using CIFS, for example, a virtual tape drive defined in MAS will be able to write directly into a directory owned and controlled by Windows 2000/XP.

Remote Synchronization

Once the MAS supports Gigabit Ethernet and TCP/IP protocol (required for CIFS and NFS), Bus-Tech intends to also offer remote data synchronization capabilities.

Remote synchronization will allow the virtual tape library to be mirrored and periodically synchronized across the Internet. Ideal for transferring data between mainframes or for disaster backup/recovery implementations, remote sync will only update data files that have changed since the last sync in order to minimize the amount of data that will have to go over the network at any point in time.

If a local MAS array should fail, critical system operations requiring virtual tape volumes can be immediately re-directed to the remote network-attached tape library. Re-construction of the local array can be performed concurrently with on-going operation.


Summary

Most System/390 installations continue to rely heavily on tape processing. Over the last few years Virtual tape systems have become popular as accounts look to reduce the costs associated with maintaining large tape libraries with thousands or tens of thousands of tapes.

High-end virtual tape systems provide expensive solutions including disk caches, tape drives, and Hieratical Storage Management software.

Bus-Tech's Mainframe Appliance for Storage takes a slightly different approach. MAS provides a lower-cost solution that attaches industry standard direct access or network-attached storage to ESCON channels and appears to the mainframe as multiple tape drives.

This approach allows the customer to use open systems storage to realize many of the returns on investment provided by much higher priced solutions and additionally provides opportunity for data exchange and remote mirroring for disaster recovery.

 



Bus-Tech, Inc.
129 Middlesex Turnpike
Burlington, MA 01803
U.S.A.: 781.272.8200
An ISO 9001

Registered Firm
Copyright 1999, Bus-Tech, Inc.