The Storage Networking Industry Association (SNIA) has produced a specification called CDMI (Cloud Data Management Interface). CDMI provides an interface specification for accessing and managing data in the Cloud. The CDMI specification was recently designated an International Standard by the Joint Technical Committee 1 (JCT 1), a joint effort of the International Organization for Standardization (ISO) and the International Electrotechnical Commission (IEC).

The standard designation by ISO has stirred up a considerable amount of interest in CDMI. I also have a lot of interest in this specification and it is a topic I will likely touch on in future posts, so I thought I would cover some background information.

Cloud storage, in general, has seen a lot of press recently. Most of the interest appears to be around the consumer based offerings for personal online storage. These isolated services fill the need for individuals, but offer very little in terms of interoperability and portability. Several services provide APIs or developer SDKs, but they are very proprietary. The more services you integrate into them and the more storage you use, the harder it becomes to move off of them. If these online storage services were to adopt a standard interface, it might be easier to migrate between them from a service integration perspective. However, moving large amounts of storage from one vendor to another would continue to be an issue.

CDMI attempts to address the first part of the problem, providing a standard interface. The surface benefits of this to both the end user and application vendor are easy to see. Vendors producing client applications could simply write their application to the standard and provide support for any service supporting the standard. End users could happily use their chosen applications on the service of their choice. While CDMI provides a mechanism for serializing data into and out of a particular vendor’s storage service implementation, physics of bandwidth and data size get in the way of this being practical.

There are a couple of features of CDMI that I particularly like:

• Capabilities

  CDMI defines a capabilities model that allows client applications to discover the capabilities of particular service. This gives client applications the ability to determine if a particular service has the features they require. The capabilities defined in the specification cover all aspects of the service, including service levels. This includes security features, throughput requirements as well as being able to define higher level features like Restore Point and Restore Time Objectives (RPO/RTO). The specification also provides a means for the service to report back to the client actual values that the service is able to achieve for these features. Having features like this provides a programmatic means for client applications to actually inspect the service level delivered.

• Metadata

  Client applications can store metadata as part of any object in CDMI. While this is not unique, the metadata can actually have an effect on an object. Most capabilities supported by the service have corresponding metadata values that can be used to trigger or modify the capability.

• Management

  CDMI provides a means of actually managing your data, not just storing and retrieving it. The specification has a concept of containers, that may be nested, acting like folders of a file system. Metadata can be placed on containers and inherited by the objects it contains.

• Non-CDMI access

  It does not always make sense to access data stored in a service through the CDMI interface. If a particular client (a web browser for example) does not know how to construct nor interpret a CDMI request, it should still be able to gain access to the content. The CDMI specification actually covers these cases, thereby standardizing how all clients access a service. This also allows one to use a CDMI enabled service as something like a web server.

The CDMI specification provides not only the basic requirements for a Cloud Storage Service, but it also includes features that provide higher level control over the stored data. These features allow for client applications that go beyond simple file storage and enables them to support the requirements of real businesses. While there is not anything that can combat the physics of infrastructure, a standard interface for data portability addresses a large set of the issues.

If you have not looked at the specification itself, head over to cdmi.sniacloud.org to check it out. The “Overview of Cloud Storage” (Section 5) is well worth the read. Another good source of information on CDMI is the Objects in Context blog which aggregates a number of CDMI sources from around the internet as well as providing original content.