Jun 23 ’09
Mainframe Storage Virtualization Extends Its Reach
Virtual Memory Operating Systems first arrived during the ’60s. Storage virtualization was pioneered on mainframes during the ’70s as virtualization concepts began to advance from the processor into storage subsystems. In the mid-70s, the original IBM 3850 Mass Storage System used virtualization to make a helical scan tape cartridge look like a disk drive to the mainframe operating system. The late ’70s witnessed the introduction of the first solid-state disk, a box of Dynamic Random Access Memory (DRAM) chips that appeared to the operating system as several rotating magnetic disks.
Virtualization reached a new level when the first Redundant Array of Inexpensive Disks (RAID) virtual disk array was announced in 1992 for mainframe systems. By combining data compression with virtualization, customers were able to store more data than the actual capacity of the array’s disk drives. Integrated virtual tape systems, which are disk arrays that serve as a front-end buffer for robotic tape libraries, first appeared in 1997 on mainframe systems. Storage Area networks (SAns) or switch virtualization appeared in the late ’90s for open systems but the seeds of this concept, called “shared DASD,” had been a mainframe staple for years. The Virtual Tape Library (VTL) concept, a derivative of the integrated VTL that had no physical tape, was introduced earlier this decade for open systems.
Virtualization techniques are most commonly used when consolidating resources. The benefits of virtualization are vast and range from a reduction in software license fees to energy savings. Storage virtualization is an abstraction layer that overlays a large number of physical storage devices. With the steady proliferation of disk arrays and tape drives in and beyond the primary data center, managing these resources is a never-ending challenge. The abstraction possible with storage virtualization can mask the unique or proprietary features of each physical storage system and provide a common way of dealing with these disparate devices, ultimately easing storage management.
Storage virtualization also provides a more consistent user interface for all storage applications. A mainframe storage array from IBM, EMC, Sun, or Hitachi Data Systems comes with its own set of storage management applications such as replication, snapshots and mirroring, but the actual storage applications are implemented differently for each system. By adding storage virtualization on top of these different products, a single set of storage applications can be initiated, so there’s no need to learn multiple user interfaces to achieve the same function; this reduces management time and costs.
Storage virtualization is widely used in file systems, SANs, switches, disk arrays, and virtual tape systems on mainframe and open systems. Users can implement storage virtualization with host software, or hardware and software appliances. Highly cost-effective solutions such as tiered storage can be simplified when based on a virtualized storage environment.
Storage virtualization is now a rapidly maturing open systems technology and is beginning to traverse both mainframe and non-mainframe systems using a common interface. Long overdue, this commonality pushes IT in the direction that “you don’t have to be a platform or operating system expert; you just need to understand storage.” This gives the customer added leverage for coping with limited personnel resources and reduces overall complexity. Knowing how, when, and where to virtualize is critical to realizing these benefits and there are several options to carefully consider.
Four Options for Storage Virtualization
Businesses have four primary options for where they can implement the virtualization architecture:
• Virtualization is provided by software installed on the host or on all application hosts.
• The virtualization engine is an out-of-band appliance or intelligent switch that passes the virtual volume descriptions to the application hosts and is operating system-independent.
• The virtualization engine is an in-band appliance located in the data path, meaning every I/O request passes through that device and is operating system-independent.
• Virtualization is provided in the storage architecture by disk array or tape controller microcode and is operating system-independent.
Comparing the Options for Storage Virtualization
Storage virtualization is extensively used in mainframes with Enterprise Systems Connection (ESCON) and Fibre Connectivity (FICON) subsystem controllers and in open systems in conjunction with SANs. An ongoing challenge the open systems SAN administrator faces is a lack of visibility regarding I/O behavior. This makes it difficult to:
• Optimize performance and utilization of storage assets
• Understand how I/O is navigating the infrastructure
• Identify where the congestion is concentrated.
Mainframes, in contrast, use control unit-based instrumentation and the host-based Resource Measurement Facility (RMF) to provide a much higher level of visibility to I/O performance. Being proprietary in the mainframe market doesn’t present the same degree of interoperability problems as it does for open systems since there are few equipment suppliers and a small number of storage management software suppliers compared to open systems. The interoperability issues involve fewer entities.
Many of the disk array and tape controller storage virtualization capabilities described in the following section are now available to mainframes and nonmainframe systems. The integrated virtual tape solution isn’t as mature for open systems as it is for mainframes; it doesn’t offer the capability for intelligent virtual volume management for tape cartridges.
Storage Subsystem Virtualization Capabilities
Virtual disk array capabilities let you:
• Create up to 32 (current) different virtual storage machines or flexible system Logical Partitions (LPARs) within a disk subsystem
• Dedicate cache, ports, back-end resource aggregation and consolidation for specific applications
• Consolidate heterogeneous storage systems and environments
• Dynamically re-allocate resources between virtual storage machines to meet changing workloads.
Integrated virtual tape (disk array and tape library) capabilities let you:
• Integrate a disk array and physical tape library. The disk array serves as cache buffer for more active data providing disk-like recall times for most accesses. As the data transparently migrates from disk to the physical tape, the logical volumes are stacked on each tape cartridge to improve utilization; typically achieving 80 percent or more utilization per cartridge.
• Decrease the number of physical tape drives as many I/O requests are serviced from the buffer
• Decrease the amount of tape media as a result of stacking multiple virtual tape volumes on a tape cartridge
• Support de-duplication and local and remote backup locations
• Support open systems
• Integrate Tier 2 (backup) and Tier 3 (long-term storage, compliance and archive) technologies into a common platform.
Virtual tape (disk only, no physical tapes) capabilities let you:
• Use a VTL, which is a disk array virtualized to appear to the operating system as tape drives and library (no physical tapes or library)
• Improve performance for the backup and recovery application because a tape mount isn’t required
• Leave existing backup software procedures intact, simplifying implementation
• Support de-duplication and local and remote backup locations via VTLs
• Use host-based virtual tape (because disk-only VTL is just emerging for mainframes).
Storage Virtualization Extends to Open Systems
Consider the open system’s evolution into virtualized storage, which essentially lagged that of the mainframe for well over a decade. For open systems, storage virtualization capabilities started on the operating system, where most of the control points were located for UNIX, Windows, and later Linux. SAN and Network-Attached Storage (NAS) have been a major initiative for more than a decade and ultimately moved much open systems storage virtualization off the server.
By 2002, the virtual tape market for open systems was gaining some momentum as a backup solution. By 2005, virtual tape, combined with de-duplication, became a popular storage initiative for many open systems storage management systems. Open systems businesses have countless, non-integrated storage management products to choose from, and storage software vendors are working to reduce the complexity and number of choices.
Organizations that have completely avoided mixed-platform storage virtualization will increasingly pay a price in terms of reduced efficiencies, greater complexity, and less IT flexibility. While the risk of staying the course may be lower and less disruptive in the near term, CIOs should carefully weigh those risks against cost, complexity, and staff availability; they should consider implementing cross-platform storage virtualization solutions as those mature. Cross-platform data storage management solutions are available that:
• Analyze reports
• Project trends
• Schedule or automate storage resources across open systems and mainframe environments.
Virtualization Improves Storage Utilization
Mainframes have consistently provided much higher levels of both storage and server utilization than open systems. Open systems storage analysis for disk capacity utilization shows that, at best, about 45 percent of available disk space is allocated. Many systems are lower. That means that about 55 percent of the installed open systems disk space is unallocated and unused. Significant resource underutilization is obviously a poor economic strategy, especially in the presence of terabyte-plus capacity disk drives.
Thin provisioning, essentially a virtualization technique that prevents over-allocation, addresses this problem and is now available on many open systems disk arrays. Not surprisingly, this capability first appeared on the mainframe in 1965 with OS/360, which enabled allocated but unused disk space to be released. Further improvements in mainframe disk storage utilization arrived with the initial implementation of Data Facility Storage Management System (DFSMS) in 1988, which became an effective policy engine for managing storage resources. By using DFSMS or similar tools, mainframe disk utilization typically reaches 80 percent or better, nearly twice that of open systems disks—significantly improving storage economics.
Because tape capacities are large and rapidly growing, now up to 1TB for native mainframe cartridges, the amount of data typically written on the tape is often less than its native capacity. When using an integrated VTL, the ability to put multiple logical volumes on a single tape cartridge greatly improves the utilization of the tape cartridge while reducing the number of cartridges. On average, tape storage benefits from a 2:1 compression ratio, essentially doubling the native cartridge capacity. Mainframe tape users with integrated tape library architectures consistently attain tape cartridge utilization levels above 80 percent.
Storage Virtualization Business Benefits
Storage virtualization offers more than just consolidation. However, with the z10 Enterprise Class (EC) mainframe, which is capable of consolidating more than 1,000 Intel servers, consolidation of the associated storage systems can’t be ignored as a substantial benefit of virtualization, especially for large enterprise data centers.
A recent survey of more than 300 technology professionals published in InformationWeek (March 23, 2009) listed the major benefits of virtualization as reducing costs, saving space via consolidation, and increased flexibility. By consolidating servers and storage, there are fewer physical assets to track and maintain, considerably reducing the administrative overhead. The reduction in maintenance costs for storage devices and servers that are removed can be significant. Users should expect to spend much less time applying updates and patches to physical servers or doing maintenance on storage devices.
Successful storage virtualization design and deployment initiatives can drive numerous business benefits:
• Improved resource consolidation, reducing facilities costs, software license costs, physical space, and energy consumption
• Reduced storage management complexity and overall costs
• Reduced human effort to manage and maintain storage
• Improved mainframe interoperability with existing “open” storage systems
• Improved device utilization levels with storage efficiency techniques such as tiered storage, integrated virtual tape, and virtual disk using thin provisioning
• Improved overall availability of the storage network by provisioning resources as needed “on the fly”
• Improved flexibility to provide or provision storage to meet changing application requirements
• Greater scalability to meet future storage needs and allow for secure migration of existing data
• Accommodating technology upgrades and replacements with minimal disruption to operations.
When properly architected, storage virtualization provides powerful, far-reaching benefits such as the ability to allocate storage resources on demand, easier integration of multiple vendors’ storage products, the ability to specifically configure storage for high availability, and reduction of the Total Cost of Ownership (TCO). The current economic downturn should accelerate virtualization efforts as “getting rid of stuff ” can slash expenses. Storage virtualization ultimately yields simplification and can significantly help the data center go green.
Choosing a storage virtualization strategy presents a challenge in the form of an over-abundance of virtualization products, resulting in considerable confusion and over-virtualization. Consider your future before you attempt to recover critical de-duplication pointers that have been compressed, encrypted, stored on virtualized Logical Units (LUNs) on a thin-provisioned RAID drive, and globally mirrored on a Wide- Area Network (WAN). Keeping the initial environments for virtualization as simple and homogeneous as possible with single vendor solutions can minimize the confusion and facilitate more orderly implementation.
The gap between management, deployment, and the amount of data to be managed continues to widen. Storage virtualization is now within the reach of everyone from mainframes to the desktop and is here to stay; it has become a compelling force in effective storage management at any level.