In this feature we look at three more buckets in its multi-tiered structure and draw some conclusions.
StorageTek is thinking about extending the object store or content-addressed storage (CAS) idea to embrace both disk and tape. Object stores are for fixed content or reference data such as mortgage and insurance policies which must be kept for many years but which are accessed infrequently. In an object store a hash-based addressing scheme is generally used to uniquely identify every piece of data or object, and both prevents duplication and ensures data integrity.
These characteristics make object stores a good fit for compliance needs. Content addressing is also commonly viewed as a way of reducing capacity as only changed data is stored instead of, as is common with backups, all data.
It is generally assumed that object stores are based on disk. However, there is no reason why they cannot encompass tape also. Access time would be slower of course, but the per/GB storage cost would be lower than disk.
There would need to be policy-based movement and management of data from disk to tape and back again to meet a business' information lifecycle (ILM) needs.
Roedel agrees, "that Object-based File Stores are similar to CAS in some value propositions but are architected to take advantage of access, performance and cost characteristics of both disk and tape under a single software management system."
With tape providing cheaper storage costs than disks this means that less active data can be stored on tape whilst the more active content can be on disk. It would be the first information lifecycle management approach to fixed content data in object stores based on two tiers of storage.
EMC has announced a tape supply partnership. Once this is in hand then it may set about extending its CAS-based Centera to include tape as part of an object store.
The new StorageTek StreamLine SL8500 library becomes available in June. It can have up to 6000 tape slots in four horse shoe-shaped rows plus a central stub and is twice as space efficient as the existing L5500 library. There can be up to 64 drives. Each one has its own fibre pipe to the outside world through a Fibre Channel port. There is a dynamically switched network inside the library.
In general StorageTek wants to see better integration of tape-based products with ILM. As part of this desire it will introduce monitoring tools covering tape libraries, tape media and SANs. The SL8500 can track tape drive status and also tape cartridge status. This allows it to detect that increased soft error rates on tape drives 1 and 2 are due to a particular tape cartridge. The sysadmin can be alerted that a particular tape cartridge, located in library slot X, needs checking.
The top row of the SL8500 is separate from the other three. It is readily conceivable that it could be removed and a lower, shorter library produced, perhaps with 250 to a thousand slots. A spokesperson said, “It could still have three robots – one per row – possibly six for great access speed. There might be just one for a deep archive.” The company says unequivocally that the SL8500 is the first in a family of products. We might expect at least one and possibly two lower capacity StreamLine library products with the lower one being a 50-200 slotter.
The SL8500 has four rack unit bays in its rear. These can be used to house disk arrays, servers and/or fabric switches. It is easy to see that the library could incorporate a disk cache and virtual tape system in one of these bays. A spokesperson said, “The backend of the library could take BladeStore arrays – mostly for caching.”
StorageTek’s next generation tape has been discussed here. Tape media will likely come from Fuji Film, using its NanoCubic technology, and also from Imation, with its Tera Angstrom technology. Terabyte capacity tape is in the foreseeable future.
We can’t see an optical storage product in StorageTek’s tiers. Roedel said, “It didn’t appear to be a strategic place for investment when we could do the same things with tape.”
The pattern of StorageTek’s development seems to be based upon having a set of storage tiers in house together with software. This provides a base from which integrated multi-tier storage services can be provided. It’s possible that a single tier vendor, such as a SATA array or a tape library vendor, would be less able to provide multi-tier integration storage services because of their limited vision and the virtually complete absence of standards in this area.
If this is true then the ILM product and service advantages lie with the multi-tier storage hardware and software vendors, such as EMC – when it gets a tape supply, HP, IBM and StorageTek. Independent software vendors such as Veritas will need to forge strong partnerships to gain the same sort of capability. Storage Resource Management vendors, such as Softek, will have to add ILM capabilities to their products or risk being left behind.
We needn’t assume that every customer will use every tier in the ILM infrastructure. Roedel thinks that, “Typical customer engagements will have three tiers.” They will be able to mix and match as their needs require.
With the continuing integration of disk and tape into an integrated backup structure then, as Roedel agrees, the position of specifically backup vendors has a question mark over them longer term. This leads to speculation that, with disk-based ILM, might we not need backup at all? Might we not snapshot copy from disk to disk to disk? But this is several years away, if it happens at all.
As StorageTek builds out its ILM infrastructure the company seems well able to claim attention from any customer considering ILM activities.