The Role Of Tape-Based Storage In Storage Area Networks.
Before we start, however, we have to address the issue of what a SAN is. To avoid writing an entirely separate article on that topic, let's summarize. For our purposes, a SAN is a separate network optimized for storage functions. It has to have two fundamental qualities to meet our definition. First, it has to provide benefit by getting storage-specific functions--large data block transfer, for example--off the LAN and onto a non-competing storage network. Second, it has to provide a high level of storage resource sharing to allow centralization of the storage management job.
Today, SANs overwhelmingly use Fibre Channel technology as the enabler, but we expect to find other technologies that will meet our basic SAN definitions working their way into the storage networking world. As they are being deployed now, SANs generally perform their functions I among servers or high performance workstations and that is likely to continue for some time. We do not see any likely near-term move toward providing widespread direct connection of desktop clients to separate, high-bandwidth, storage consolidating networks.
SANs Are Allowing Consolidation Of Backup
To see what roles tape is likely to have in future SANs, let's look at how it's being used in the SANs that are working today. Tape-based systems are central to SANs today because the majority of SAN installations are built to make it easier and faster to back data up-and tape is still the dominant technology for backup. In some ways, SANs represent a critical next stage of an ongoing backup consolidation trend. The trend starts with IT departments backing up server data with a tape drive on each machine, using manual processes to change tapes, kick-off backup jobs, and dig tapes out of a vault to restore files. When there's too much data to make this practical, people consolidate their backup and automate the process, usually by bringing data over the LAN to a centralized backup server and connect that to an automated tape library. As data increases, the system scales by adding additional data backup zones, each served by its own server and its own library. As data reaches a critical mass--and this is happen ing in a huge number of enterprises--network traffic is affected negatively and the overhead required to manage many different distributed storage systems becomes excessive.
This is the point at which IT departments need to find a different solution to keep moving ahead and it is where Fibre Channel SANs are providing an elegant solution. In today's typical SAN, Fibre Channel Host-Bus Adapters (HBAs) connect each of the servers to a separate Storage Area network that includes a single, centralized tape storage library. Data is now moved directly from each server over the SAN with connectivity created by high-bandwidth switches. The first benefit is that all that backup data transfer instantly disappears from the LAN and the IT staff manages only one library instead of several. The total number of tape drives required for the backup is also reduced because Fibre Channel, unlike SCSI connections, lets a storage target, like a drive, be dynamically shifted from one source to another. Practically, that means that I can back up many servers with few drives--a real advantage--and finally, adding additional storage resources (like additional drives or libraries) becomes easier because F ibre Channel SANs can add new addresses without shutting down the system (Fig 1).
SAN backup will continue to evolve past the initial LAN-free stage and it's starting to now. The next major step will be widespread availability of serverless backup. At its simplest, this utility simply allows data to be moved from disk to tape libraries directly through the switched fabric without using the processing resources of the individual servers to move the data. Instead, an active agent in the fabric moves the data while the servers simply initiate the action and arbitrate the use of shared resources (Fig 2). The active agent can be located anywhere in the fabric of the SAN--ADIC has just introduced tape libraries that incorporate the agent directly so that they become, in a sense, an intelligent data moving resource.
So what kind of role will tape systems play in the world of SAN backup? They will, from all the evidence that we see today, continue to be crucial resources. The SAN environment, however, will require slightly different tape system configurations. We can expect, for example, to see fewer enterprise servers backed up with stand-alone tape drives. More will be backed up with automated libraries and we expect the size of those libraries to increase as data is consolidated. The ADIC Scalar models, for example, integrate intelligence into the library itself in order to offload some of the current server functions. We are seeing a strong demand for SAN libraries that grow and change configurations easily, for as single libraries serve dynamic, high-growth, centralized data pools, flexibility and scalability will provide extremely high value.
SANs Will Use Tape For Tiered Storage
If SANs did nothing more than help consolidate backup, they would continue to drive use of automated tape storage systems, but SANs will also move toward allowing disk-based storage to be shared among multiple servers. This is happening in only a few applications now, but it is likely to grow rapidly. The earliest adopters of SANs for sharing disk-based storage simply use ports on the switches to divide a common array into separate volumes--not too different from what people do with SCSI-attached disk. This is changing, though, with the advent of true SAN file sharing systems that let multiple users actually share common data down to the individual file level. The initial applications of this technology share very large files among users in collaborative workflow settings--applications like digital effects for motion pictures or high-resolution graphics. The next major application is allowing many hosts to share single copies of many smaller files in dynamic file-serving environments, like web sites and e-com merce. In each case, the effect of the shared SAN storage is to minimize disk requirements and eliminate unnecessary duplication and transfer of files because the SAN bandwidth allows all connected hosts to access centralized files at local disk speeds (Fig 3).
What role will tape storage play in these new SAN applications? I believe that the most likely result will be another increase in tape use, this time an increase in the role of tape as a tool for storing and serving data rather than simply backing it up. The reasons are fairly straightforward. Data in the enterprise is doubling every year and SANs are helping accelerate an overall trend toward consolidating data into larger centralized pools. Shared disk and SAN file sharing will only accelerate that trend. With more data in centralized locations, it will become more important and easier to manage data more intelligently--to treat different categories of data in different ways depending on what it is being used for.
High performance RAID systems will continue to provide critical functionality in SANs, but just as SCSI-based arrays, they won't solve all of an enterprise's storage needs. Enterprise RAID systems will continue to be great at providing rapid access to data for active processing and providing protection against single disk crashes. Yet their cost will continue to be high and they will still need other storage media--tape primarily--in order to provide protection against large-scale disasters and most classes of data loss because of human error. The idea that some (RAID manufacturers mostly) have floated again recently about eliminating backup to removable media with a scheme that stores all data on more and more replicated RAID arrays isn't economically practical for end users today. It doesn't look to get any more practical with the advent of SANs. High performance disk--SCSI or SAN--isn't going to become inexpensive, removable, and self-managing in any near-term time frame.
In the non-SAN world we're already seeing data growth and new connectivity options bring tape into a more active storage role. Standard SCSI disk all across the enterprise is clogged with data that is essentially static and managing it is becoming a major problem everywhere. One possible solution is ADIC's StorNext, a NAS appliance that uses low-cost, high-capacity tape storage to let people get static data off their active disks, but still keep it available automatically and transparently to end users.
In the SAN world, a similar use of tape as part of a consolidated storage strategy is highly attractive and, because the SAN environment will be dealing with very high data concentrations, the ideal system will put only the data that needs rapid access and high speed processing on the relatively expensive active SAN disk. For very high volume data and for data that is accessed less frequently and needed less quickly, tape is likely to play a major role. What makes sense is to have disk and tape storage available over the switched fabric and to automate migration of data between those resources in a way that is transparent and easy to administer. Serverless data transfer built-in to the devices will keep even large-scale transfer from interfering with normal server functions. The key to making this work in the client-server world will be to make the administration as easy as possible. The approach that we favor is to combine file sharing for active disk, automated tape storage, and a policy-based migration sys tem within an easily administered SAN appliance. It will provide flexible data and file sharing directly for servers and performance workstations and will serve files to clients through the SAN-attached servers. We expect to see the first stages of products built around this architecture available before the end of this year.
Virtualization 01 SAN Data
The trends we have just described being created by SANs can also be seen as part of two great paradigm shifts taking place in network architectures, ones that have been gaining attention from many in the storage and networking industries. In the first, SANs are placing information--data itself in its stored form--at the center of the enterprise, in a sense, replacing servers as the key network component. This shift is well underway with existing SAN applications and will rapidly accelerate as serverless backup becomes widespread. In it, data storage in all its forms--including disk and tape--stands at the center of the enterprise. Networks will be built around the data, the way it is stored, and the variety of ways in which it is moved, protected, and accessed.
In the second, only just beginning to take shape, the data is in the process of becoming abstracted from its location--RAM, local disk, shared disk, or tape-- becoming, instead, a shared, virtualized commodity. This second shift will mean that the location of particular files will be selected transparently to users and administrators through a variety of automated systems and different storage devices based on the kind of data and the demands for its use. As SANs drive this virtualized storage forward, the varied uses for data--its tiers of utility--will continue to mandate a wide variety of storage mediums. We believe that all the signs point to tape storage systems continuing to play a crucial role in the way that SANs help organizations store, protect, and use data, their most valuable asset.
Dave Uvelli is the executive director of software marketing at ADIC
|Printer friendly Cite/link Email Feedback|
|Title Annotation:||Industry Trend or Event|
|Publication:||Computer Technology Review|
|Date:||Aug 1, 2000|
|Previous Article:||MP3 Brings High Capacity Disk Drives Into The Home.|
|Next Article:||NetVault Reincarnated ... With Bakbone.|