Photo: Wong Tran
With block storage, data is organised as an array of unrelated blocks and access to the blocks is directly controlled by a host. With file and network-attached storage (NAS), data is similarly organised as unrelated blocks but the onboard file system places data on the disks. External systems provide access to the files within the onboard file system.
In object storage, the approach is to provide application-centric data storage, access and management. Virtual containers are used to encapsulate data, their attributes, metadata, and object IDs or keys.
"The easiest way to store a large amount of data is using a "bucket" to keep different objects within. Object storage doesn't care about file structures."
Object storage led the way towards cloud storage, which offers a number of benefits. It has almost limitless scalability designed to manage the explosion of unstructured data, and it is capable of scaling across multiple physical locations, regardless of distance. It also has advanced, policy-driven data management based on attributes and metadata, while its flexible access methods support traditional object storage models and new Web-based application architectures.
He mentioned DropBox as an example. It is globally distributed but uses object storage in the backend. Another example is the Amazon S3 cloud where each storage file is an object. Such object storage typically has characteristics such as a single flat namespace for objects, location-independent addressing, and "constant time" read performance. "It also has unlimited storage, and is auto-configuring," added Tran.
>> Go to report on Plenary Session.
>> Go to report on Business Technology Stream.
Sign up for Computerworld eNewsletters.