Object-based storage is a concept that was developed to help provide a solution to the ever-growing data storage needs that have accompanied the IT explosion since the late twentieth century. It acts as a counterpart to block-based storage, allowing large sets of files to be grouped together and to move the processing power for those files away from server and workstation CPUs and closer to the storage itself. This processing power is utilized to assist in the implementation of such features as fine- grained security policies, space management, and data abstraction.
Since object-based storage is not addressed in blocks, like most of the storage used in everyday workstation and server environments, the object storage device (OSD) interface requires some way to find out how to address the data it contains. Objects are the individual pieces of data that are stored in a cloud storage system. They are composed of parts: an object data component, which is usually a file that is designated to be stored in the cloud storage system, and an object metadata component, which is a collection of values that describe object qualities. The OSD interface uses object IDs as a unique identifier for the combination of data and metadata that comprise each of the objects.
Along with all the files that each object contains is an associated set of metadata that can be used to describe the data component of a specific object, and classify it or define relationships with other objects. This metadata is an extensible set of attributes that is either implemented by the OSD directly for some of the more common attributes or interpreted by higher-level storage systems that the OSD uses for its persistent storage.
A binary large object, or BLOB, is a collected set of binary data that is stored as a single, discrete entity in a database management system. By gathering this binary data into larger collections, database administrators are able to better copy large amounts of data between databases with significantly reduced risk of error correction or data filtering.
Policies are similar to metadata in that they are attributes associated with the object. The difference is that policy tags contain information that is associated with a particular security mechanism.
One of the primary uses of object-based storage is the practice of working with replicas. Replicas are essentially copies of one large set of data, often associated with a virtual hard drive or virtual machine. They are used to both increase availability and reduce the amount of risk associated with keeping a large amount of data in one location.
Replicas are good candidates for object-based storage for several reasons:
They are large datasets that require a copying mechanism that can run efficiently without requiring expensive error correction or filtering.
They do not affect user performance SLAs if they are faced with I/O latency, which is often associated with object-based storage.
Published on Tue 28 February 2012 by Frank Lazenby in Computer Science with tag(s): storage objects