Storage Virtualization--Strategic For SANs.

So old, it's new

The concept of virtualization has been used in many facets of the computing industry dating back to the late 1960s when CP/67, a very early operating system developed by IBM, initially appeared. CP/67 evolved into the VM (Virtual Machine) operating system. VM allowed users to run several operating systems on a single processor machine. Later, in the early 1970s, Virtual Storage appeared in large-scale operating systems giving programs the illusion that they had the capability of addressing far more main storage (memory) than the machine actually contained. By the mid-1970s, virtualization was well accepted by users of various operating systems. Virtual concepts have since been used in many architectures and designs including: Virtual Address Extension (VAX); Virtual Circuit (VC); Virtual Machine (VM); MVS (Multiple Virtual Storage and became OS/390 and now is z/OS); Virtual Memory System (VMS); Virtual Private Network (VPN); Virtual Reality (VR); Virtual Storage Access Method (VSAM); Virtual Storage Extended (VSE; was DOS/VSE); Virtual Tape Subsystem (VTS); and Virtual Telecommunications Access Method (VTAM).

One of the predominant current applications of virtualization concepts is in the storage industry. We can define storage virtualization as the presentation of a single interface to users that provides a logical view of data or storage devices different from the way data or storage is physically configured. With virtualization techniques, the user does not need to know how storage devices are configured, where they are located, or what their capacity limit is. Thus both the logical and physical characteristics are separated from each other allowing the application to access the logical image while minimizing any potential disadvantages of the underlying device or storage subsystems. By separating logical and physical characteristics, physical devices can be added, upgraded, or replaced without disrupting the application or server's availability. Hardware or software, and most often both, are used to accomplish storage virtualization. The storage networking movement is redefining the values of virtualization.

Virtualization first appeared in the data storage industry in 1974 when IBM delivered the 3850 Mass Storage System (MSS). The MSS implemented virtualization by making the physical storage, actually robotically retrieved helical scan tapes, appear as rotating magnetic disk drives to the operating system. Virtualization next appeared in the late 1970s with the arrival of the first Solid State Disk (SSD) from Storage Technology (now StorageTek) Corp. The SSD appeared logically as an ultra-fast rotating disk drive but physically consisted of no moving parts, containing only DRAM (Dynamic Random Access Memory) chips. In the early 1980s, several disk storage suppliers offered disk subsystems that had multiple logical disk images on a single physical disk drive. RAID and volume management software were emerging as virtualization technologies.

In 1994, the industry's first virtual disk array, originally named Iceberg, was delivered by StorageTek providing the disk industry with the first implementation of outboard data compression, log structured files, and RAID 6+ architecture. The system's logical view of the storage appeared to be one of a much larger storage subsystem containing more volumes than the physical storage subsystem actually offered.

By 1997, storage virtualization had finally reached the tape industry as several suppliers began shipping various implementations of virtual tape with IBM and StorageTek becoming the primary providers. Virtual tape implemented disk arrays that actually appeared to the operating system as tape drives serving as a buffer between online disk and nearline or tape library storage. Host application data that was re-referenced within a few days after being moved to nearline could actually be retrieved from virtual tape (physically the disk array) without requiring a tape robotic mount. This addresses the long-standing issue of the time it takes to get to the first byte of data on a tape or removable storage library by accessing more highly used data from the virtual buffer. If the data in the virtual tape system is not referenced in a given period of time, it is physically moved to a robotic nearline storage system. Again, the application saw a logical view of storage that was different than the physical storage subsystem.

Storage virtualization is one of the three new trends in the Storage Networking industry along with SANs and NAS as the 21st century begins. SANs have and will continue to take storage virtualization to new levels and offer far more potential benefit than previous virtualization implementations. Virtualization is very strategic for SANs.

VERITAS, among others, has delivered a suite of continuously improving virtualization SAN management solutions. With a logical view of storage, every attached server can act as if it has its own dedicated physical storage while, in reality, the storage is not physically dedicated to any single server. Zoning (access is controlled by restricting connections between designated ports), LUN masking (a filter that allows access to specified storage resources), and partitioning (defining logical storage subsystems) are good examples of SAN virtualization available today. SAN virtualization architectures provide a tighter integration of the storage hierarchy and are aimed at moving data seamlessly across the various layers of the storage hierarchy while at the same time providing transparency to the users. The race to build a complete SAN makes a key attribute of virtual storage systems moving storage intelligence, data management, and control functions outboard freeing server cycles while providing the needed effi ciencies of centralized storage management.

The NAS market is also beginning to deploy virtualization techniques. Tricord Systems recently announced Lunar Flare NAS, becoming the first appliance to utilize the company's Illumina software, which aggregates the appliances to create a cluster that is managed as a single resource or virtual storage pool. Illumina-enabled appliances provide seamless growth and continuous access to content. Lunar Flare NAS makes it simplified storage management and makes it non-disruptive to grow storage resources. Adding an appliance to a cluster takes minimal time and doesn't interrupt clients' access to data.

Key functions of storage network virtualization include:

1. Pooling

* Storage devices appear as one large storage pool.

* Provides for a single point for consistent policy definition and management control.

* Applies to disk and tape storage systems.

2. Connectivity

* Connects any (heterogeneous) server to the storage pool, any-to-any connectivity.

3. Upgrades

* Enables non-disruptive capacity expansion, critical when storage is growing between 60-100% annually for most users.

4. Security

* Permits access to storage by only those applications that have access rights.

5. Outboard functionality

* Move data seamlessly and transparently across layers of the storage hierarchy.

* Server-less backup and recovery.

* Accelerates the demand for true data sharing (heterogeneous servers sharing a single copy of data).

* Suggests an outboard HSM (Hierarchical Storage Management) system that can handle blocks and files possibly residing in a storage domain manager.

Software is key to exploiting the benefits of virtual storage systems. Partitioning, LUN management, and zoning all use software interfaces to activate and change the functionality parameters of the SAN. The arrival of server-less backup and recovery, disk to tape file transfer, snapshot copies, forensic backups, and non-disruptive database extractions are further implementations that take advantage of storage virtualization techniques.

Virtualization techniques are now beginning to appear and deliver significant functionality for storage networking. Specifically, the network based approach to storage virtualization implements the virtualization functionality embedded within network firmware or software attached to the networking devices such as switches, routers, appliances, or domain directors. This seems like the most logical approach to control data traffic and storage management for SANs becoming the foundation for the future SAN manager. In this example, it becomes possible for a storage administrator to implement a single storage policy for all attached storage.

These approaches are emerging as either symmetric, where both the control information and data go through the same path, and asymmetric, where there are separate paths for control and data flow. By using switches and routers (which sit in the data path), the storage and virtualization agents do not have to run on the attached hosts avoiding the security issues of host-based approaches while offering better performance. Availability can be improved by offering switch, router, or director redundancy supporting multi-path fail-over.

Storage virtualization is being implemented with a variety of schemes. Presently, no standards exist for virtualization. As is most often the case in the IT industry, de-facto standards usually win. It is important to evaluate storage virtualization techniques strategically by identifying more complete solutions that will evolve to combine both disk and tape storage management rather than just focusing on singular storage management problems.

Storage virtualization techniques are fundamental to the evolution of the SAN and toward our continued march toward building and delivering the information utility. Virtualization enables viewing things differently than they really are. Vision is not seeing things as they are; rather it is seeing things as they will be. Applying vision to virtualization can create disruptive changes. Where is virtualization in your future?