OSD: The Holy Grail of storage?

By Kenneth Samarra

Imagine a storage device that can grow dynamically as needed, without interruption of service. Envision this device connecting to any operating system, with simultaneous access to the same data. Imagine, also, that the device can manage itself.

So far, it has been as unattainable as the Holy Grail. But the Object-based Storage Device (OSD) may be close at hand.

Today, there are two ways to grow storage. You can backup all your data, add more space, then restore the data. This process can easily take hours for terabytes of data. The other solution is to stage disk data onto tape with hierarchical storage management (HSM) software. However, with the HSM approach, older files take much longer to retrieve because they are on a slower device (such as a tape library), and eventually your cache will outgrow your available disk space. In contrast, OSD will allow for disk storage to be expanded on the fly just by adding more drives.

In an Object-based Storage Device (OSD) model, the file system storage component is moved from the operating system to the storage device.
Click here to enlarge image

To date, users have not been able to construct a truly heterogeneous storage area network (SAN), because there is no standard file system format across platforms. Even when attached to the same operating system, a locking mechanism (e.g., clustering) needs to be implemented to coordinate access to a shared disk. With OSD, you can have it all-multiple operating systems in coexistence.

Storage management can cost more than the hardware itself. In the OSD model, all metadata resides on the device. This will allow for self-management and a more effective use of cache and pre-fetching. Not only will storage take care of growing itself, it will also perform off-line backups and RAID (or RAIT for tape).

How will OSD accomplish all this? The answer is intelligent drives. Storage devices have had on-board microprocessors for years. However, this power has been significantly underutilized. By giving intelligence to the drives themselves, the OSD model allows for a complete device abstraction from the host operating system(s). This is accomplished by breaking out the file system storage component from the operating system and moving it onto the storage device (see diagram).

Sector and logical block addressing (LBA) will be accomplished by the device itself, leaving the operating system to communicate via a new OSD interface. With this mechanism, the host operating system only needs to send simple commands (e.g., create, open, read, write, close, etc.) to access the storage device. The operating system no longer accesses disk drives on a logical block level.

And OSD is not just limited to disk. It will work with devices such as tape libraries as well, so that an operating system such as Windows will not think that it owns tape devices exclusively. Essentially, everything on the SAN will be an object, capable of communicating with all other similar objects.

OSD has moved from the NASD Project of NSIC and SNIA into the ANSI T10 ANSI Committee for approval as an industry standard.

The Holy Grail of storage is on the horizon, but the quest is not yet complete. In order for the storage industry to fully attain the vision of OSD, many industry leaders will have to adopt this new standard.

A team of experts from Hewlett-Packard, IBM, Quantum, Seagate, StorageTek, and Carnegie Mellon University are working to not only define the technical specification, but also to evangelize OSD's merits. But it is not enough for the hardware vendors to collaborate. If OSD is to succeed, the operating system vendors must also cooperate. After all, they are going to be giving up some control over these new devices. Middleware suppliers such as Legato and Veritas will have to make modifications to their software. Database developers will have to re-tune their engines to accommodate the new structure. Likewise, host bus adapter vendors will have to modify their drivers to accommodate OSD devices.

It seems like a daunting task, yet it is attainable. The simplicity of this solution is much more elegant than the current practice of writing a tangle of workarounds to compensate for out-dated storage design. SANs don't need to be hard to implement.

For more information on OSD, visit the ANSI T10 Committee's site at: ftp://ftp.t10.org/t10/drafts/osd/osd-r01.pdf.

Kenneth Samarra (ken@kairos.com) is CEO of Kairos Information Technology, and is a member of the Fibre Channel Standards Committee. He also acts as a senior consultant to StorageTek and other vendors in the storage industry.

This article was originally published on September 01, 2000