Original URL: http://www.theregister.co.uk/2010/10/29/caringo_castor_v5/

Caringo administers dose of CAStor oil

Constipated file systems need grunt and strain no more

By Chris Mellor

Posted in Storage, 29th October 2010 13:51 GMT

Caringo, founded by the Centera technology inventors, has reformulated its CAStor oil for the fifth time, adding named objects and more granular multi-tenancy.

Caringo was founded by the people who devised the FilePool technology which EMC bought and turned into its Centera product line. CAStor is a content-addressed storage (CAS) technology running, unlike Centera, on commodity X86 hardware nodes that are clustered and self-managing, and automatically check data integrity and provide replication-based, not RAID-based, protection. Dell is OEM'ing the technology in its DX6000 object storage platform.

The content sky is getting populated with more object storage products aiming to displace filesystem-based approaches to cloud storage on grounds of far more efficient object location than by running across a filesystem tree. As the number of objects scales to a billions and beyond file system approaches become more and more disadvantageous.

That's the story being put by Caringo, CleverSafe, HDS, and newcomer Scality, aiming to give filers an object lesson in cloud storage.

Version 5.0 of CAStor introduces user-named objects which scan be used alongside the system-provided Unique User IDs (UUID). These named objects are stored in abstractions called buckets, and buckets for each main user or tenant of a CAStor cloud are stored in domains. Security policies can be applied at domain, bucket or individual object level.

Caringo's named objects, buckets and domains can be thought of as being very loosely, equivalent to files, folders and filesystems layered onto an object storage base with the named object infrastructure operating in parallel with the object infrastructure using UUIDS for objects. The firm says named objects are suited for applications requiring a self-generated or symbolic name to store and retrieve a file or object.

Domains can be unique to a department, division or company and thus, like HDS' updated HCP enable a tenant's storage to be broken down into sub-tenancies. Caringo says Amazon's S3 can only have one domain per user (tenant) whereas CAStor can provide many.

CAStor v5.0 has the concept of caching high-activity objects in a node's DRAM. Where HCP uses HDI nodes to cache content in edge, on-ramp devices, CAStor 5 caches content across the content store cluster heart, and says its operation in this sense is akin to a content delivery network. As demand for an object changes, CAStor automatically manages the overall cache to increase or decrease the number of cached copies throughout the storage cluster.

Caringo says CAStor is a cloud storage system, deployable internally as a private cloud or by a service provider as a public cloud. Users access their data by a username/password combo and access control lists (ACLs), with this data held as a named object, and users given access to single or multiple objects only, up to buckets and domains.

The company hopes, no doubt, that its CAStor object storage will help file systems kick the bucket, in the cloud at any rate. ®