CAS: storage for fixed content; getting the most out of all your information - Storage Networking - content addressed storage

Computer Technology Review, Oct, 2002 by Tom Heiser

Single-instance store: Centera maintains only one copy of a content object and a mirror. If, for example, an application attempts to store the same content object for twenty different clients, for all twenty users, the metadata within each user's CDF would be different, but the object itself will still be stored only once. Because mathematically only one content address exists for a single piece of content. The result is unprecedented capacity-savings and simplified management. Just one of the reasons why the TCO associated with Centera is so attractive.

Centera's implementation of content addressing protects the system from unauthorized access: The content address does not map to a data path, file name, or data type. A user has no "account" on Centera and there is no way to browse. The only way to access content on Centera is through an application, which must have the CDF's 27-character content-derived alphanumeric address.

A single CDF can reference multiple digital assets, such as a collection of files and folders.

The Centera CAS Solution

Centera employs a unique RAIN (redundant array of independent nodes) architecture. Each node consists of a CPU, large capacity disk drives, and runs the Centera operating environment, CentraStar--the heart and soul of Centera. The CentraStar operating environment automates content placement in Centera and runs on every node. A fully populated six-foot cabinet provides approximately 10 terabytes of mirrored content storage (20 terabytes of raw storage). Multiple cabinets can be configured into a single environment enabling Centera to scale up to more than a petabyte of capacity. Centera also can be configured so that content is automatically replicated from one Centera to a remote Centera while still being mirrored locally. This ability to store content in different geographical areas has important disaster recovery benefits. For example, in the event of an outage, replication of content would allow an organization whose local IT resources no longer function to quickly access data at the remote site with virtu ally no down time.

CentraStar is designed to handle minor hiccups and major issues without human intervention. It allows Centera to self-configure, self-heal, and self-manage. For example, adding capacity is a straightforward process requiring minimal time and effort, and is totally non-disruptive. Centera simply "auto-discovers" and uses new capacity as it becomes available. Centera configures and manages itself. At installation, the system administrator need only specify the desired number of nodes to be used as access nodes, based on bandwidth requirements, and configure their IP addresses. CentraStar handles the rest, automatically configuring the internal IP addresses of Centera's nodes and switches. CentraStar also takes responsibility for where content is physically stored within Centera. New content goes to the two nodes that are least busy and have available space--two nodes because every stored content object has a content mirror.