Object storage is one of the three main protocols to access persistent external storage. It complements block and file by offering a highly scalable, data-agnostic platform to store and retrieve binary objects or files over standard web-based HTTP protocols.
Object storage has gained in popularity in recent years as the demand for low-cost online backup and archive has increased. It isn’t reserved for secondary data, though, and can also used to store large volumes of primary unstructured data.
The number of object storage startups continues to rise as solutions are brought to the market as software-only implementations.
Object storage doesn’t have the same rigorous latency requirements of block-based systems but is sensitive to data throughput instead. Performance in this instance is achieved by deploying multiple scale-out resilient nodes, either as virtual machines or physical servers.
The opportunity in this part of the market is therefore much more intense, with fewer barriers to entry for newcomers.
The differentiators between suppliers are features (with the addition of multi-protocol support) or high performance. Outside of that, competition is based purely on a $/TB model, which means prices fall over time.
One interesting development is the move towards releasing solutions as open source. In this part of the market at least, storage is definitely commoditising.
Swarm is Caringo’s software-defined storage object solution. It can be deployed natively on standard x86 servers, within virtual machines or even the public cloud. Each node integrates into a collection of storage nodes, providing features such as multi-tenancy and flexible billing and auditing.
Swarm supports multiple site capability using a feature called parallel replication. It supports the S3 protocol as well as authentication through traditional platforms such as LDAP and Active Directory.
Caringo recently added extended search capabilities, and its FileFly product allows Swarm to be used as a large file archive store.
HyperStore from Cloudian is another software-based object storage platform. HyperStore can be deployed either as an appliance or software-only on the customer’s own hardware.
Cloudian offers three model types. The 1500 appliance delivers 12 drives in 1U of rack space with up to 96TB of capacity, based on 8TB drives. The FL3000 series is a 3U chassis of eight storage nodes, each with up to either 90TB (FL3020) or 180TB (FL3050) of capacity.
Extra expansion shelves of 60 drives in 4U can be added to an initial FL3000 node chassis. HyperStore is a scale-out node-based architecture that claims 100% compatibility with the AWS S3 protocol.
The object solution from DDN (Data Direct Networks) is called WOS and can be deployed either as an appliance or as a software-only solution.
The WOS7000 appliance offers high-performance nodes, based on a 4U chassis and space for up to 60 drives (maximum capacity 480TB using 8TB drives). The WOS capacity appliance supports up to 96 SATA SMR drives in a single 5U chassis for a maximum capacity of 768TB (8TB drives).
A single WOS cluster can scale to a maximum of 256 nodes, with up to 32 clusters in a single namespace. Protocol support includes S3, Swift and standard NAS (NFSv3, NFSv4, CFS, SMB 2.1 and SMB 2.2).
HGST acquired Amplidata’s object storage technology in March 2015 and has used the technology as the basis of its Active Archive System.
The platform comes in two models: the base rack SA1000 and a fully populated SA7000. The SA1000 offers 672TB of raw (417TB usable) capacity, scaling to 4.7PB raw (2.9PB usable) in a single rack.
Throughput performance is rated at up to 3.5Gbit/s with a full rack configuration. Active Archive supports AWS S3 and standard NAS protocols for connectivity (NFS/SMB), and can be deployed across multiple geographically dispersed locations.
A relatively new entrant to the object storage market, NooBaa, like many other solutions presented here, is a scale-out object store based on multiple nodes or servers running the Linux operating system.
NooBaa can run on VMs or physical machines, with each node contributing local storage into a pool that is then represented out as an object store. One node is designated as a managing cluster and administers the node membership of other servers in the cluster. Pricing is based on capacity configured within each cluster of servers.
Another software-defined storage product, OpenIO is based on an architecture of many scale-out nodes. The platform itself is open source, with the code freely available on GitHub.
The software can be deployed within a VM or directly on hardware, with support for RHEL, CentOS or Fedora Linux operating systems. OpenIO has been delivering its object platform since 2008 and claims to have more than 10 petabytes of capacity under its storage and tens of billions of objects.
It says its solution scales to exabytes of capacity in a single cluster, and offers advanced features such as policy-based management, versioning, compression and data deduplication.
Ceph (or to be more accurate Inktank, Ceph’s developer) was acquired by Red Hat in April 2014. Although the software remains open source, Red Hat packages and sells a commercial version of Ceph, marketed as Red Hat Ceph Storage.
The Ceph platform is a software-only product based on multiple storage nodes and a technology called Rados (reliable autonomic distributed object store) that lays out and manages data across multiple clusters.
Ceph supports S3, Swift and native object protocols as well as providing file and block storage offerings. It is cited as the most popular storage platform for OpenStack distributions, particularly for Cinder (block) support.
Scality’s Ring software is a software-based solution that can be deployed on standard hardware (using either CentOS/RHEL, Ubuntu or Debian operating systems), including that from technology partners Dell and HPE, both of which resell the Ring platform.
Ring supports native multi-protocol access, including AWS S3, Scality’s native REST API, NFS, SMB, FUSE and OpenStack protocols (Swift, Glance, Cinder and Manila). A single Scality Ring can be deployed on one site, or across multiple geographically distant locations as a stretched Ring.
More on object storage
- In the first of a two-part series we run the rule over the big six storage suppliers’ solutions for object storage, an increasingly popular protocol for unstructured data.
- Object storage is a rising star in data storage, especially for cloud and web use. But what are the pros and cons of cloud object storage vs building in-house?
Scality doesn’t quote any specific platform metrics, other than claiming to support configurations based on thousands of nodes and unlimited numbers of objects. High-profile deployments include the 30PB of capacity used by the US Los Alamos National Laboratory.
At heart a tape hardware company, SpectraLogic has done something interesting with its release of Black Pearl, first announced in October 2013. The platform acts as an object gateway to Spectra’s tape libraries, storing and retrieving data on tape using the LTFS standard.
Black Pearl uses a superset of the S3 protocol, with extensions to cope with the issues of managing data on tape media. The Black Pearl appliance gateways sit in front of tape libraries, acting as a cache. The 4U appliance claims a 1Gbit/s throughput rate.
SwiftStack is a startup that offers a managed release of the OpenStack Swift object storage project. Founded in 2011, the company takes the Swift open source software and adds management and operational functionality. Pricing is based on capacity consumed.
Release 4.0 of SwiftStack introduces features such as enhanced load balancing across nodes in a cluster, and metadata search capabilities. Hardware requirements are based on minimum recommendations (such as 1GB of RAM per node for each TB of disk) rather than being offered as an appliance.
Zadara has recently expanded its cloud-focused storage offerings with Zios, a scale-out object storage platform. Zios can be installed either on-premises or accessed through Zadara’s public cloud, deployed globally in conjunction with major cloud providers like Amazon Web Services.
Zios offers data replication using mirroring, with the option to keep either two or three copies of each object. Protocol support includes AWS S3 API and Swift. Zios is part of a wider offering from Zadara that includes block-based and file-based storage, delivered by its VPSA (virtual private storage array) platform.