Tape Virtualization At StorageTek.

Virtualization is a hot topic these days, with many storage companies excitedly announcing its addition to their products. In its simplest form, virtualization fools the host into seeing the storage device it expects and needs--even though the reality might be quite different. Some of the most recent examples of virtualization are occurring in SAN environments, where software and hardware products create virtual volumes across a number of storage devices.

However, virtualization in the tape arena has been around a decade. It was based on mainframe storage, or the lack thereof-storage as a resource was scarce in mainframe environments and administrators needed to squeeze every bit of storage they could out of available tape. Elements of tape virtualization impact media, libraries, and tape drives, and are used for redundancy and/or performance gains. Library virtualization might include storage consolidation, fail-over, and mirroring/striping. Tape drive and media virtualization primarily presents particular devices to applications with specific drivers or device needs, including device model emulation, type emulation, and interface abstraction. As Quantum's Rory Bolt puts it, tape drive virtualization decouples applications from technologies.

Virtualized tape libraries in particular can provide huge benefits, including a much lower total cost of storage represented by capacity per cubic inch and ease of management. They can also offer very high data and system availability, effective resource utilization, and application independence.

One of the earliest practitioners of virtualization, StorageTek, is working to develop and expand virtualization products as the company seeks to turn around its financial for tunes. According to the company, benefits of tape virtualization include improved performance, storage management reduction, and increased hardware and environmental efficiencies

Improved performance. Tape data sets are written to a disk buffer that allows users to quickly access an active file instead of waiting on tape restore times. Faster tape mounts and more consistent mount times, faster batch processing, backup, and restore capabilities also reduce data management times.

Storage management reduction. Virtualization minimizes manual tasks associated with tape management and allows resources to be deployed elsewhere. This is particularly true with automated dynamic volume/LUN allocation, since manual LUN masking is a primary administrative task/nightmare in non-virtualized environments. Administrators can also save time by managing multiple platforms from one GUI interface.

Hardware and environmental efficiencies. Tape virtualization allows for total automation, resulting in significantly less wasted tape space and slots. It can also reduce tape drive numbers and associated maintenance costs because it uses much more of the tape capacity than backup programs and other applications using reserve allocations.

Two of StorageTek's virtualization products are its nearline library, Virtual Storage Manager (VSM) and its network-attached pooling device, StorageNet 6000.

Virtual Storage Manager

The Virtual Storage Manager provides nearline storage by using virtual (appropriately enough) tape technology and automation to optimize hatch processing, tape utilization, and cartridge capacity. VSM has dual ACS, media duplexing, and a fault-tolerant architecture including robotics and disk buffer storage. Offering up to 64 virtual tape drives per system and a nearly unlimited number of virtual cartridges, the device provides both device and media emulation. VSM first shipped 98Q4, and now StorageTek has over 1000 VSM systems in customer sites.

The VSM assigns newly created files to a disk buffer where they can remain immediately available for a period of time. When a disk buffer reaches a customer-assigned capacity, internal algorithms assign active files to an automatic migration queue onto the virtual tape drives. Any operating system or application that can access a tape drive will work with the VSM.

Customers are primarily using the VSM for batch type and journaling operations--applications that generate a large amount of data but not at a high rate of speed. This is the case for most nearline systems, which do not handle high-speed processing needs well. (Many DBAs would love to use nearline systems in their data warehouses, but since nearline tends to increase query time five to seven times from online-based database transactions, the warehouses require much more expensive RAID-based storage.) But for applications such as the above, or with application-based backups, nearline systems such as VSM are excellent solutions.

The predominant cause of using tape instead of hard drive is cost. Even with library and tape subsystems, an order of magnitude less makes them extremely cost-effective. This is true not only in purchase price but in maintenance: when archival data goes around spinning on a hard drive, it costs both electricity and heat. But it costs very little to have data sitting on a cartridge. And as applications are notorious for squandering tape, the VSM saves money by causing them to use much greater cartridge capacity.

StorageNet 6000

The StorageNet 6000 (SN6000) is a storage domain manager (SDM) that provides tape pooling in a multi-server environment. Scaled for different levels of connectivity and redundancy, the SN6000 series provides device virtualization, a single point of management, and any-to-any Fibre Channel (FC) port connectivity.

StorageTek classifies the SN6000 as network-attached storage. It incorporates the Virtual Transport Manager (VTM), which enables users to share their tape device assets across all of their Unix and NT hosts, as well as across multiple backup applications. Cost savings are key here, at up to half the cost of RAID and other disk-based storage solutions. With VTM, SN6000 can consolidate a number of tape devices across many application servers, so hosts can use new storage device technology without changing their existing device drivers.

SN6000 devices can be connected, though arguably few customers actually need to do that--their top-end device already has 16 interfaces, each of which can be configured to present 16 virtual tape devices. In terms of optimal performance, most sites will configure from four to eight in each interface. Still, that's a lot of virtual devices for one 6000.

The product shipped in October 2000 and is presently in a production capacity at about 30 customer sites. StorageTek plans further additions to the SN6000, including disk storage and OS/390 environments. Although it terms the SN6000 a network-attached storage solution, StorageTek is pushing the SN6000 as an example of its hoped-for leadership in heterogeneous SAN environments.