Paul Speciale, VP Product Management at Scality.
Data&StorageAsean: In your own words - What is Object Storage?
Paul Speciale: Object storage is a model for storage that stores and manages data as key/value pairs in a flat namespace. All storage systems have the concept of an identifier for a piece of data, such as a physical disk address in a block-based system such as a SAN, or a file name in a NAS or file system. In object storage, the keys are the identifiers for the values (data objects) stored in the object storage system. The values are indeed the actual data payloads, or to fit the terminology the "objects" in the object storage system. Hence, we have a way to store and retrieve the objects through keys - think of this as a valet parking system where you hand over your ticket (the identifier), and what is retrieved is your car (the value). You don't have to know the physical or even logical location of the object, only the key, for accessing the data. The flat namespace part of object storage refers to the lack of complex naming hierarchies such as in a file system, where we have the concept of folders, sub-folders and files - so you actually have to know how to navigate to and find files. In a flat namespace such as in an object store we essentially remove the complexity and overhead of managing this complex naming structure, mainly to make this simpler and more scalable.
Data&StorageAsean: What are the major use cases for Object Storage Solutions?
Paul Speciale: Object storage is useful for use-cases where:
The number of objects will scale beyond the traditional limits of a file system. Commonly file systems work well to 100's of thousands or perhaps millions of files (and much less than that in one folder is typically) advised.
Applications manage access to the data.
In cloud storage services where access to data over http/REST based access protocols are useful for accessing data remotely over the internet is required. With block based and file based storage, local access was the only efficient deployment model.
With these criteria in mind, Object Storage is now effectively deployed in massive scale data archives, private and public clouds, and large media repositories where the file count can be very large and the access may require cloud based streaming access.
Data&StorageAsean: Should object storage be on premise or in the cloud?
Paul Speciale: It can be both, enterprises are creating private storage clouds based on object storage for the scalability reasons outlined above, while service providers and cloud services use it for scaling access and capacity to massive levels, driven by increasing numbers of end-users.
Data&StorageAsean: What are some of the key features people should look for when choosing an object storage solution?
Paul Speciale: The system should primarily be reliable and durable in terms of protecting data. Next, the system should be designed to be highly available even when failure events occur with components such as servers and disk drives, but also with the software managing the system. Finally, it should be proven to deliver scalability to billions of objects and higher, to support tens to hundreds of petabytes of storage capacity.
Data&StorageAsean: Is there anything unique about your company’s object storage offering(s)?
Paul Speciale: Yes, the Scality RING has several differentiators:
First it offers a combined object storage and file storage solution, to help companies as they transition to the new style of cloud applications. This means that both legacy apps and new style apps can coexist in one system to make management easier.
The Scality RING has been proven to scale to tens of billions of objects in one system, this is due to our unique parallel design with no single points of failure or bottlenecks.
The Scality RING is suitable for a wide range of data sizes, from small objects and files such as in email/webmail use-cases, to massive object sizes in media for high-resolution 4K video files. No other object storage system is as efficient and performant for this wide range of data.