In fact object storage is not exactly new, but it is a proposition mushrooming with a number of new vendors. And, it does require a fair bit of explanation and as a consequence has still got an air of novelty about it.
Revision update The 1st version of this post was published on Nov. 23. Since then several comments were made and as a result a notice was placed to alert readers that it was up for revision. Thanks for the suggestions, input etc.

What is behind the ‘object storage’ label?

The ‘object’ in the name alludes to the fact that things behind the scenes are located in different ways to how things are traditionally done.

What is traditional?

The world of computing eventually realised that it was helpful that a part of the operating system could be dedicated to organising how data was stored and retrieved. This part became known as the File System. An index, catalog or directory etc. takes care of knowing where chunks of files reside. File systems became more sophisticated because of the needs to deal with different devices and access control, plus a variety of useful utilities and reporting. And a function of the file system was farmed out into database software systems because they needed to be directly in control of data and improve performance.

Why different?

One group of computer scientists quickly finds flaws with the status quo and looks at doing things differently. In the case of databases these were only the answer to some of the needs because they brought with them their own set of restrictions. Object storage is organising how data is stored obviating the need for a File System, and is different from a database. The motivation for doing things differently lies in the reality of the computing schema combined with what problems object based storage is trying to fix. A earlier parallel was CAS [1] footnote.

What is better?

This is where the vendors agree and disagree – see How they differ. The essential benefits to object storage is that it is designed to serve up data faster and better. What is stored and retrieved is an object, so it can be a variety of formats. Object storage is also better at scaling into vast numbers of objects and nodes. This means slicing up storage-compute and storage-data and networking the whole thing. This is an architectural different design from putting many disks next to a powerful storage controller which is then networked.

How they differ?

There is consensus in this segment that object storage is a common cause. But these vendors are also competitors and inevitably each implementation has got its own benefits. Accordingly the vendors distinguish themselves in numerous ways: implementation, features, vertical focus, security, data protection, appliance based vs. software etc.

What about metadata?

Object storage metadata provides information about the location and other attributes about the data itself. Each object data implementation is unique to each vendor and its metadata is thus proprietary. Conversely, IT customers do not need to know how a File System stores its meta data. The similar case is made why customers need not know how object storage solutions go about their metadata.

Object storage is born

Object storage is a way of fixing the inhibitions or outright problems of less desirable alternatives. But the vendors proposing object storage have a challenge in front of them because it takes a bit of explaining. Hence, the need to promote a common cause. The next thing they do is to take a deep breath in order to explain how they distinguish themselves from the other offerings in this category. Storage systems once delivered very high profit margins, but these are coming down. Object storage has the potential to accelerate this trend.

Recently Q4 tradeshows in Europe have been the platform for such efforts. Here are some vendor observations.

Amplidata

Amplidata’s A20 Storage Nodes are 1U high, with 20TB of storage and preloaded with Bit Dynamics agents to cope with scale, self-optimisation, secure data. The nodes stack up to provide full systems and the protection regime is based on Erasure Coding to supersede the wanting RAID-X concept.

Caringo

Caringo. The CAStor solution is software only, but has a hardware OEM in the Dell DX6000. A node can be created from a USB key and subsequent nodes automatically grows the cluster. Primary access from applications is via HTTP/REST.

Cleversafe

Cleversafe can be deployed as software only or running on Cleversafe’s own hardware solution. The software takes care of scale & distribution of encrypted unstructured data with monitoring and optimisation.

Scality

Scality’s RING Organic Storage and RAIN (Redundant Array of Independent Nodes) distributes data on X86 servers and deals with different data structures, scale, distribution of data & control and cloud. The initial customers are recruited from hosting, service and cloud providers. It articulates its benefits in terms of TCO analysis.

These are just four vendors who compete in Object Storage solutions. All four are accessed by interfaces such as HTTP/REST.
This is neither an exhaustive list of vendors nor their offerings. The initial customer cases come from specific vertical industries and IT service providers feature prominently. What these all share is a need to deliver IT via IT Clouds. As Cloud spreads then more verticals will explore Object Storage.

Footnote:

A parallel was Content Addressed Storage (CAS) which was an appliance to serve up data. CAS also introduced a new way of protection.