Object storage and object stores

What is object storage and what are object stores?

Object storage is a method of managing data storage in discrete units, known as objects. An object store is a platform where data analytics software can execute queries on objects.

With infinite capacity, object storage was once viewed as inexpensive storage for backup and recovery functions. As developers moved to the cloud and analytics grew in enterprises, the use cases for S3 and object storage have grown due to its ability to handle massive data volumes and data sets.

How does object storage work?

Object storage keeps blocks of data together in one unit and assigns a unique identifier with rich metadata to each object for easier searching anywhere within a distributed data pool, whether on-prem or across multiple network systems and geographic areas.

Object storage eliminates the hierarchical structure used by other systems by using a flat address space. This way, it can be scaled up or down easily, accommodating rapid expansions and contractions as storage workloads vary.

In addition, to make updates to stored data, object storage systems save new copies of the entire asset and not just the small portions that were modified. Because of that, object storage has more durable data than that within other systems.


Unfortunately, a heavier metadata load demands more overhead and slows down data modification processes. This being the case, traditional object storage was better suited to data that needs less frequent access, such as:

  • backup processing for archive retrieval following a disaster or data loss
  • static web content, such as photos, historical documents, music, etc.
  • applications that write just once or that are just read and/or not updated frequently.


Object storage vs block storage vs file storage

Data storage comes in three mode categories that serve data access needs differently: object, block and file. Which method to choose depends on several key characteristics that make use cases better suited for one versus the other.

As the oldest and simplest data storage method, file-based storage puts data as a single piece of information inside a folder, much like a paper-based system. Because the stored data can be accessed with a simple pathway, it lends itself well to shared file management. But because of its hierarchical filing structure, finding one specific piece of information can be tedious at best.

Block-based storage divides data into units that can be spread across a system for more efficient storage. Block storage, however, limits the metadata assigned to each data block to a unique identifying address. Because of this, blocks don’t demand much overhead and can be nimbler and more efficient than other systems. This leanness makes for extremely high performance for frequently changing data, such as transactional data.

By contrast, object storage adds much more metadata to each file than other storage systems. It can include contextual and customisable information, such as application details, data protection levels and other information that might pertain to retention policies or other attributes. This allows users to deploy a wide variety of analytics on large quantities of unstructured data. And as an additional benefit, object storage eliminates the hierarchical structure of other modes, allowing for higher scalability and more broadly distributed access.

How does object storage work with Kubernetes?

To understand how Kubernetes and object storage intersect, it’s important to understand that, fundamentally, data growth drives technologies into obsolescence. As a result, the rapid and continuous increase in data volume is leading to an even faster pace for technology development cycles.

Designed specifically for an environment of rapid change, Kubernetes has become a dominant force in how operators manage computing, networking and storage infrastructure. As the default platform for Kubernetes, object storage offers the elasticity, scalability and resilience that Kubernetes needs to power its build/package/deploy framework. Using object storage, Kubernetes allows operators to handle everything, from provisioning to volume placement, at scale and with reduced overhead.

And when applications run in containers, object storage maintains the state of these applications. This is critical for Kubernetes because it requires the object storage itself to run in the container in order to be able to manage infrastructure automation. By using object storage, Kubernetes can therefore orchestrate infrastructure in a stateless, portable environment.


HPE object storage solutions

As use cases for object storage expanded, so did HPE’s need to offer multiple solutions.

HPE Solutions for Scality is the preferred solution for general-purpose object storage and includes:

  • on-prem alternative to public cloud storage from massive scale down to single node edge
  • media and ready-to-share data repository
  • next tier storage for data offload, e.g. Splunk SmartStore, medical imaging and HPC archive
  • simple enterprise backup target storage. 


HPE Ezmeral Data Fabric is the preferred solution for analytics, AI and ML workloads that need high performance and high scalability.  Some examples of these use cases include:  analytics, research, Internet of Things (IoT) and business insights. HPE Ezmeral Data Fabric is the first solution to unify file, object, NoSQL databases and streams into a single unified data infrastructure and file system across on-prem, multicloud and edge environments.  This enables apps and users to directly access trusted data no matter where it resides.