A reel life saver
A reel life saver
It's not sexy technology; it's slow and frustrating; and it's more to think about compared with the disks we know and love. But it could save your business one day.
Love it or hate it, tape is as much a part of corporate IT strategy as computers are-particularly as ever-increasing numbers of servers store ever-increasing volumes of data.
With system integrity absolutely essential, moving data from those servers within available backup windows has become a primary measure of a company's ability to react and recover in the event of a disaster.
If tape is the vehicle for preserving your precious data, multi-drive tape libraries are the freeway to get it where it needs to go. Ranging from basic two-tape models up to monstrous robotic libraries whose thousands of tapes fill up a room, there are tape libraries in sizes to suit every environment.
Choosing the right library has been a relatively simple endeavour in the past: divide the amount of data you have to back up by the speed of the tape drives times the number of drives in the library, and you had a good idea of your actual requirements. Quadruple those as a rule of thumb, and you'll have a sense of what you need to get you through a few years' growth.
Bigger, better, faster
The laws of physics put tremendous demands on digital backup tapes, something that's forced tape vendors to stagger their long-term technology roadmaps as they work to resolve the multiple issues of friction, pulling force, data density and error correction that are involved in producing a reliable tape.
Constant progress has ensured a more or less predictable growth in tape capacities, with last year's launch of next-generation LTO technology pushing tape capacities to 200GB of uncompressed data and speeds of around 80MB/second. The next generation will sit around 400GB uncompressed, with 800GB and 1TB the targets by the end of the decade (in 2002, HP crammed 1TB onto an experimental tape in the laboratory, but is still working out how to shorten the 18 hours it took to write).
Real innovation in the tape sector comes from the new designs and capabilities that vendors are building into their libraries. The number of tapes supported within the libraries has already broadened enough that it's not really necessary to fit more into a chassis; that increase will come through steady increases in tape size. Another key issue in tape library performance, the bandwidth connecting it to the network, has also been well handled with the proliferation of direct Fibre Channel connections that link them straight into storage area networks.
This year, the biggest changes in tape libraries will come as vendors reduce physical library footprints and reduce the number of drives within individual libraries in an effort to provide customers with more flexible tape solutions.
Many of the changes are predicated around improving telecommunications links between corporate branch offices: as remote office bandwidth increases through use of DSL and dedicated IP WAN connections, it's possible to move dedicated tape libraries out of the field back into head office. This allows companies to manage their entire backup regime centrally, rather than relying on a remote receptionist to push the right tape into the backup drive every night.
Whereas past tape libraries were islands unto themselves, coming libraries will take a modular approach that allows customers to buy small devices, then add extras as storage requirements grow. In this paradigm, the libraries automatically recognise and interact with each other-a move that vendors believe will let customers start small and grow their tape libraries incrementally. This has them, predictably, smelling blood in the potentially lucrative small to medium enterprise market.
"There is definitely a movement towards fewer, larger libraries and at the same time sharing fewer, better tape drives across smaller servers," says Rob Nieboer, storage strategist with tape library and disk giant StorageTek. "Rather than having [lots] of little tape drives that might be used for an hour or two every day with relatively little reliability, we'll see fewer, larger libraries with fewer, better tape drives being shared across servers. We can add capacity non-disruptively by putting additional capacity alongside each other. The target for us is less than an hour of downtime a year."
Increasing the duty cycle of tape libraries requires considerable work: tape drives are mechanical devices, after all, and mechanical devices wear the more they're used. In high-use environments, this fact may make the modular approach even more appealing since failed drives or libraries could be readily swapped to meet backup windows.
ILM helps tape find itself again
If the physical layout of tape libraries is becoming more flexible, so too is the role that vendors envision for tape libraries in light of the industry shift to embrace information lifecycle management (ILM).
ILM, in which data is tracked and moved between classes of physical storage devices according to its relative importance and currency, necessarily involves the pairing of tape libraries with disk - which has often been mooted as a possible replacement for tape.
The reality of ILM lies somewhere inbetween, and it's becoming increasingly clear that tape libraries will be most effective when paired with inexpensive arrays of ATA-attached disk.
So far, those ATA arrays have come in the form of near-line storage arrays such as StorageTek's DX100. Within the next few years, however, tape libraries will increasingly be designed to expect that a higher-speed disk buffer be available to insulate the tape from the servers it's backing up.
This change will have a significant effect on information management: rather than having to back up every single file to tape that has to be archived, the introduction of a disk buffer will allow data to live on that buffer for a set period of time-and then be archived off to tape. In this configuration, non-essential information such as temporary files can be expired while it's still on disk, leaving the tape as the ultimate repository for the smaller amounts of critical information.
In the first half of this year, StorageTek will set a trend by introducing a tape library that incorporates its own ATA disk buffer. This buffer-which will scale from 17TB to as much as 500TB over time-will look to the rest of the network just like a tape drive, meaning that applications will believe the data has been backed up. The library can take responsibility for the data from there, allowing it to complete the ILM cycle without having to involve the servers generating the data.
This approach also makes the most of the snapshotting capabilities built into modern operating systems, allowing customers to keep a recent snapshot online for faster recovery. "The big disadvantage with tape has always been the time to do backups and restores," says Ian Selway, storage product manager with HP. "If you've invested hundreds of thousands in disk arrays to run applications, you don't want them taken offline while you're doing a backup. Introducing staging disk in the middle of the tape backup lifecycle means you can restore from disk in a much shorter period of time."
ILM has fast become a key element of the push for increased corporate governance, and tape libraries are set to change in response. This year, IBM will introduce libraries that offer a WORM (write once read many) capability for permanent archiving of data-allowing customers to use tape to create permanent, unalterable long-term records of emails, documents and other critical information.
As market pressures continue to raise the bar on customers' retention of corporate data, tape will play an increasingly strategic role. Even for customers with extant libraries, coming technologies are worth another look as ILM becomes an integral part of information management strategies.