Key Criteria for Evaluating Object Storagev3.0

1. Summary

The object storage market is more active than ever. Object storage is very popular with developers, and just about every IT organization is adopting it in one way or another. Not only is it common in the public cloud, especially with large organizations, but hybrid infrastructures and large on-premises installations are growing as well.

There are several reasons for the increasing demand. Today, a growing number of applications need to store data safely and access it from everywhere, from multiple applications and devices concurrently. Moreover, high-performance computing applications—like big data analytics and AI—are also big consumers of object storage, and the new flash media types introduced in high-performance object storage systems strengthen the case for object storage generalization among high-performance, high-value workloads. This means that the key characteristics of enterprise object stores have changed, with much more attention paid to performance, ease of deployment, security, federation capabilities, and multi-tenancy than in the past. In fact, users are taking advantage of their experience with the public cloud to build on-premises infrastructures that can support as many workloads as possible while keeping costs under control. In addition, edge use cases are becoming more prevalent. Tiny object storage installations at the edge, serving small Kubernetes clusters and IoT infrastructure are surfacing. This introduces several new factors that will soon become relevant, such as managing a fleet of edge devices, edge-to-core data replication, and distributed object storage services.

Most of the success of object storage comes from its use in the public cloud. Amazon AWS S3 API is the de facto standard, and all software and hardware vendors adopted this interface for their solutions. Other APIs, like Microsoft Azure Blob, are now available on some products, but they are not as common.

In any case, API standardization and the resulting enhanced compatibility enable users to build hybrid and multicloud infrastructures more confidently and at a reasonable cost. Even better, some vendors have proposed multicloud data controllers aimed at virtualizing access to multiple object stores on the back end for better control over data placement, governance, and cost. Standardization on the S3 API simplified the access layer and helped accelerate product differentiation in the back end. The S3 API has now been adopted by multiple storage systems, including primary scale-up storage systems, and products are very much differentiated in terms of scalability, performance, and features (see Figure 1).

Figure 1: Various Types of Object Stores Compared

Modern scale-out object stores are becoming increasingly specialized. While some products can be considered general-purpose storage to serve as many use cases as possible, these solutions don’t excel in any specific vertical. Others are becoming more focused on specific applications delivering the best return on investment (ROI), with features and add-ons designed for those workloads. However, they often lack important enterprise features. Whether to choose a general-purpose approach over a more focused one depends on business needs, including the amount of data that has to be stored and the number of applications the object store has to serve.

As with any other IT infrastructure component, security is a key area that must be carefully addressed. With more and more applications using object stores to save their data persistently and permanently, security is crucial. For example, object storage is often used both as a primary backup target and for disaster recovery purposes. S3 offers several security features on the front end, but to provide strong end-to-end protection against data loss and malicious attacks, object stores must include additional security features and characteristics. Ransomware protection is an emerging topic: While object stores already provide object immutability features, organizations will expect a more proactive stance toward ransomware detection and attack prevention.

Finally, the manageability and ease of use of object storage are expected to improve: Some solutions already benefit from AI/ML-based management and analytics improvements that are already available in block and file storage systems.

Object storage usually has a very long lifespan compared to other storage systems in the infrastructure, and the market landscape is very complex. Before deciding which solution is better, users should consider whether the object store will be relatively small and serve one or a few applications or whether it will become a fundamental infrastructure service that will store data for a large number of applications, services, and other infrastructure components. This decision will have a strong impact on both data management and your deployment model.