The business benefits of virtual tape
Combining hard disk and tape results in lower cost of ownership, reduced management, improved asset utilization, and better data access.
A virtual tape storage management (VTSM) system, which combines a disk-array front-end (or buffer) with a physical tape back-end, has significant benefits in automated tape library environments. Virtual tape is targeted at system/390 mainframe environments, where inefficient media use is an issue. In Unix and NT environments, tape is primarily limited to backup.
VTSM is not a solution for all applications across the enterprise. For example, VTSM is not a good candidate for backup and large sequential batch jobs. These applications, which typically create large volumes of data with long data transfers, are better served by physical tape. Also, they already make efficient use of the physical media and they typically have a low probability of recall. The virtual tape system`s disk buffer is best allocated to data sets with a high probability of recall that can benefit from the high-speed response of hard disks.
One application that is well suited for VTSM is batch processing, which is characterized by both media inefficiency and a high probability of recall. Batch jobs typically create small tape data sets with one data set per cartridge. Virtual tape addresses media usage, significantly improving data storage efficiency. Also, virtual tape`s high-speed hard-disk buffer can significantly improve batch-processing time and overall reduce the batch window.
Data sets with a high probability of short-term retention and expiration, such as temporary data sets, are also good candidates for virtual tape. These data sets can expire and be flushed from the storage subsystem electronically, eliminating unnecessary back-end migration/recall/recycle overhead and minimizing media and library-slot consumption. A disk buffer`s fast recall also significantly reduces processing time.
VTSM is also a potential solution for small data sets archived for long-term retention. These data sets don`t typically require virtual tape`s recall performance capabilities, but by passing them through the disk buffer and migrating them to the back-end physical storage repository, they make better use of data storage space, which translates into cost savings for users.
A VTSM system accommodates large numbers of virtual volumes and can scale up to large disk buffer capacities. The bigger the buffer, the longer a virtual volume resides in the buffer and the greater the probability of recall from the buffer rather than the back-end physical tape. The batch window shrinks as the number of recalls from the buffer increases. Up to a 50% reduction is possible. By reducing the batch window, the availability of on-line applications is extended.
Since simultaneous read/write/migration/ recall/recycle operations will be the norm, the disk buffer engine supports the processing demand without compromising response times and service-level requirements.
The key components for performance are the number of I/O data paths (both between the host and disk buffer and between the buffer and back-end physical storage), the number of virtual and physical tape drives, and the disk buffer I/O throughput.
Another important consideration is intelligent data management. Many users want to reduce manual management effort, while others want to actively manage the data. The system should accommodate intelligent grouping and stacking (i.e., by application or expiration date), duplexing (for business resumption plans), and migration/recall policies (i.e., immediate migration, pre-staging, media preferencing, etc.). These functions can be set up in a management class structure for simplified storage management.
Data compression can also add value to a VTSM system. Storing compressed data increases the effective capacity of the disk buffer and improves migration/recall response times. Faster migrations/recalls also occupy physical back-end storage devices for less time, freeing them up for other processing tasks. Transferring compressed data remotely also more efficiently uses communications resources.
The back-end physical repository, on the other hand, is scalable in terms of both capacity and performance. This is necessary to accommodate new application growth and changing service level requirements.
The VTSM system and back-end physical repository should support multiple drive and media types because no one technology is best suited to deliver the cost/performance requirements for the diverse application portfolio of the typical enterprise.
To minimize additional costs, a VTSM system should leverage existing asset investments and be compatible with existing tape automation products, including physical tape drives, libraries, slots, media, and library control software. And back-end physical devices should be shareable between virtual and non-virtual applications.
A VTSM system can free up a significant number of slots and pieces of media. The net result depends on the average usage of existing media, but if you convert to new tape technology, the reduction in library slots can range from 200:1 to 2,000:1.
For many sites, footprint reduction is the primary goal. For others, the freed up storage capacity can be used to accommodate application growth, "open" platform and network backups, and new business applications such as document management, microfilm replacement, video, etc.
The business benefits of tape automation and virtual tape include:
Improved asset usage, including physical tape drives, libraries, and media.
Lower cost of ownership through more efficient asset usage, reduced physical resource requirements, and reclaimed CPU cycles.
Reduced management resources through the buffering, stacking, migration/recall, and recycling of data sets--all outboard to the host processor.
Extended investment protection by integrating existing tape automation environments; leveraging existing tape drives, libraries and media; and freeing up media and slots.
Improved customer service through reduced batch windows and improved service level delivery.
Improved data access through greater availability of on-line applications by leveraging tape automation over a broader base of applications and enabling automated access to more archived data.
A typical configuration consists of an S/390 processor attached to both a physical library and a virtual tape management subsystem. The disk buffer contains virtual volumes, virtual tape drives, and front-end and back-end data paths. Migration, recall, and recycling occur between the disk buffer and the physical storage repository, without any host intervention.
Dave Osekavage is manager of enterprise Nearline marketing at Storage Technology Corp. (www.storagetek.com), in Louisville, CO.