Create a Relational Distributed Object Store

ion allows for comparatively naïve users and administrators, as the object store takes care of the detail of how and where user data is stored, protected, geo-replicated, deduplicated, versioned, garbage collected, and so forth. Abstraction also greatly simplifies application development and deployment, as evidenced by the plethora of start-ups, who are able to quickly get a worldwide service up, running and generating revenue by leveraging a public object storage service in unprecedented timeframes. Likewise, we see where limited compute platforms, such as smartphones and tablets, are able to have full access to a universe of data despite zero support for a single storage protocol. Such devices have no support for the traditional storage protocols [Fibre Channel, Small Computer System Interface (SCSI), Network File System (NFS), and Common Internet File System (CIFS)] made popular in the last century for local-area-network (LAN) attached disk(13). Instead, both simple and complex data types are served by nothing more than the basic Web Hypertext Transfer Protocol (HTTP). Contributions This paper describes a mechanism for adding a relational layer on top of the object storage layer, without destroying the simplicity gained through the abstractions for which object stores are noted. Our goal is for users to be able to manage relations between data in a fashion similar to a relational database. However, since unstructured data is intrinsically different from structured data, it is necessary that we provide a substrate for defining and describing such relation. Further, we envision creating a mechanism by which users can query an RDOS in a manner similar to the way they query a relational database today for similar benefit. Document Organization The remainder of the document is organized as follows. “Unstructured Data” defines unstructured data and touches on some of the associated challenges this data type brings. Metadata is essential to the challenge of creating a relational layer on top of unstructured data. “Metadata” defines metadata, explains why it is of such value, and identifies its potential to transform unstructured data from an undifferentiated data mass to a highly correlated data set that provides genuine value to the business. “Hitachi Content Platform” describes HCP in its present form and “Relational Distributed Object Store” describes the ideal of extending it to be a relational distributed object store and understanding the business value it provides. “Related Work” and “Conclusions” provide a summary of the topics covered in this paper.