Magic Quadrant for Distributed File Systems and Object StorageLeaders are Dell EMC, IBM and Scality.
This is a Press Release edited by StorageNewsletter.com on 2016.10.25
The report, Magic Quadrant for Distributed File Systems and Object Storage, was published on 20 October, 2016, and written by Julia Palmer, Arun Chandrasekaran, Raj Bala, analysts at Gartner, Inc.
Bimodal approaches that demand infrastructure agility and scalability are stimulating market adoption of scale-out file and object storage products. This research helps I&O leaders assess the key attributes, vision and execution prowess of distributed file systems and object storage market vendors.
Strategic Planning Assumption
By 2021, more than 80% of enterprise data will be stored in scale-out storage systems in enterprise and cloud data centers, up from 30% today.
Scale-out storage systems are growing fast and becoming a popular platform to tackle the unabated growth of unstructured data. With data growth exceeding 40% year over year in many enterprises, infrastructure and operations (I&O) leaders are looking for extensible storage products that can address an increasing number of use cases with lower acquisition and operational costs.
Enterprises are demanding features and capabilities prevalent in big data cloud infrastructures, such as self-healing and ease of management. Software-defined storage (SDS), deployed on commodity hardware, is emerging as a threat to external controller-based (ECB) storage arrays in environments with a steep growth of unstructured data. New and established storage vendors are continuing to develop scalable storage clustered file systems and object storage products to address cost and scalability limitations in traditional, scale-up storage environments.
Gartner defines distributed file systems and object storage as software and hardware solutions that offer object and/or scale-out file technology to address requirements for unstructured data growth and based on 'shared nothing architecture.' A shared nothing architecture is a distributed computing architecture in which each node is independent and self-sufficient, and there is no single point of contention across the system (see the Acronym Key and Glossary Terms section).
Distributed file system storage uses a single parallel file system to cluster multiple storage nodes together, presenting a single namespace and storage pool to provide high bandwidth for multiple hosts in parallel. Data is distributed over multiple nodes in the cluster to deliver data availability and resilience in a self-healing manner, and to provide high throughput and capacity linearly.
Object storage refers to devices and software that house data in structures called "objects," and serve clients via RESTful HTTP APIs, such as Amazon Simple Storage Service (S3) and OpenStack Swift.
Vendor Strengths and Cautions
Caringo was established in 2005 and is privately held. Based in Austin, TX, Caringo's main offering is Swarm, an object storage product, which was originally launched in 2006 as CAStor and is now in its eighth version of software release. The product has successfully focused on the needs of customers in sectors such as healthcare that require a scalable and secure solution for storing data for regulatory compliance purposes with built-in search capabilities. Caringo Swarm consists of redundant array of independent nodes (RAIN), where each node is capable of performing all operations. There are no file systems or centralized metadata controllers, which enables a simple, resilient and scalable architecture that has a flat peer-to-peer structure. Caringo has a patented power savings technology called Darkive, which spins down idle disks for energy savings and extension of hardware life. The Swarm product can be deployed either bare-metal or as a VM image on VMware ESXi or Microsoft Azure. Dell was formerly an OEM partner of Caringo with a Dell-branded appliance that was discontinued in 2013. While Dell continues to be a reseller partner of Caringo's today, with the EMC acquisition, this partnership looks uncertain.
Most of Caringo's customers are based in the U.S. and Western Europe. The vendor relies on a direct and inside sales force to stimulate market demand for its products. Caringo has no direct presence in the AsiaPac region or Latin America, and is dependent on channel partners to service and support those geographies.
- Caringo Swarm is a mature object storage system that delivers competitive security and compliance features, including legal hold and WORM capabilities, to ensure data immutability.
- The pricing of Caringo Swarm is all-inclusive, competitive and flexible. Caringo offers both a capacity-based perpetual licensing model as well as subscription-based licensing with the option of small, incremental additions of capacity.
- Caringo's FileFly software, based on an OEM offering from Moonwalk, provides a simple and non-disruptive way to migrate older files from existing NetApp and Windows systems, while maintaining metadata and ownership rights.
- Caringo is a privately-held company that is yet to be profitable. Moreover, it has largely been self-funded in the past several years, which has limited its ability to aggressively invest in its business.
- Caringo has a weak partnership with server OEMs that limits its enterprise appeal due to a lack of wide reference architectures and integrated support from the hardware vendors.
- While Caringo has been in existence for more than a decade, its deployments tend to be smaller in size by capacity when compared to many of its peers.
Cloudian is a Silicon Valley-based start-up that develops HyperStore, an Amazon S3-compatible object storage platform that has been available since March 2011. HyperStore is delivered as a turnkey integrated appliance or as a software-only product that can run on industry-standard hardware. HyperStore offers on-premises storage and data management, as well as replication, erasure coding and multitenancy services, and tiering to the Amazon Web Services (AWS) cloud. A core part of HyperStore's appeal is its close compatibility with Amazon S3, which means that applications and tools designed to be used with Amazon S3 will likely work with Cloudian's object storage platform. The vendor has made a number of notable updates to HyperStore, such as an improved management UI with distributed cluster management features, and has enabled crossregion replication. Cloudian has introduced rolling upgrades and added HyperStore Connect for Network File System (NFS) and Server Message Block (SMB) support.
One of the initial use cases for HyperStore was the storage and backup as a service for cloud service providers (CSPs). However, the vendor has shifted from focusing on service providers in Japan to a more mainstream enterprise market worldwide, and now has a mixture of customers from more-diverse verticals, geographies and sizes.
In the last 12 months, Cloudian has reinforced its relationship with OEM hardware partners (in particular Lenovo), as well as released new Cloudian hardware appliances for both entry-level and multi petabyte-scale markets.
- HyperStore maintains closer compatibility with the Amazon S3 API than much of its competition.
- Cloudian's multitenancy and QoS features resonate well with large enterprises and CSPs.
- HyperStore appliances provide robust, modular, high-density and resilient hardware options for a wide range of end-user capacities.
- Gartner clients outside of Japan and the U.S. are rarely evaluating HyperStore due to Cloudian's limited market traction outside of those geographies.
- Cloudian's marketing efforts are, in some cases, ahead of the reality of HyperStore's success in terms of traction in the market, size of deployments and level of S3 API compatibility.
- A cornerstone of Cloudian's vision includes hybrid cloud storage capabilities that involve data tiering to public cloud services such as Amazon S3, but the market adoption of such features is negligibly low.
DataDirect Networks (DDN) is a privately-held company based in Santa Clara, CA. The vendor offers Web Object Scaler (WOS), an object storage product that supports a native web servicesinterface in addition to providing Amazon S3 and Swift compatibility. While DDN is a mature storage supplier to the HPC market, much of its focus with WOS is outside of this space. HPC customers often deploy DDN's WOS-Bridge product that allows WOS to be used as an archive tier for data that no longer requires high-performance primary storage.
WOS can be configured with policies to use replication and/or erasure coding depending on the data protection and RTO requirements of the workload. DDN typically recommends that a customer use a combination of replication and erasure coding for purposes of site failover in order to avoid elongated rebuild times when significant failure occurs. However, there are costs and RTO that should be evaluated by workload to avoid creating duplicate copies of data. DDN offers WOS Access, a NAS gateway that offers NFS and SMB on the front end to communicate with the WOS object storage over a web service protocol on the back end. WOS Access can be deployed at remote office locations to provide NFS and SMB access in a hub-and-spoke deployment model with WOS on the back end. WOS ArchiveDirector is a file archiving gateway that provides retention management by policy to prevent accidental or intentional deletion of content, along with compression and deduplication. WOS supports LDAP and Active Directory integration for import and management of users within WOS.
DDN has sold WOS as a turnkey hardware appliance since 2008 and has more recently shifted to allowing WOS to be deployed as software with broad hardware support. As such, DDN offers reference architectures of WOS with Dell, Hewlett Packard Enterprise (HPE), Supermicro and the Open Compute Project (OCP).
- DDN has a heritage in the HPC market and has the resources in place to serve these customers well with WOS.
- Enterprises can select from a variety of WOS hardware and software deployment options that suits infrastructure, cost and management preferences.
- DDN offers seamless integration between WOS and its parallel file system appliances, GRIDScaler and EXAScaler, for handling warm and cold datasets between these environments transparently.
- WOS prioritizes performance over security features, and lacks server-side encryption and HTTP authentication.
- WOS's S3 API compatibility, which uses Apache Hbase, is implemented on an architecture that does not scale well beyond a single site.
- Some reference customers that currently use WOS rated DDN poorly in terms of delivering on sales promises in addition to solving challenges with respect to deploying and operating complex WOS-based infrastructures.
Dell EMC has two products that are included in this research: Dell EMC Isilon and Dell EMC Elastic Cloud Storage (ECS). With Isilon, a mature scale-out NAS product with deployment across many enterprises and service provider verticals and geographies, end users consolidate large datasets onto one distributed file system for high scalability and throughput. ECS is a platform that supports object, NFS and Hadoop Distributed File System (HDFS), with the ability to store billions of objects for cloud-native applications.
The Isilon OneFS OS is on its eighth generation of software release and has more than 7,000 customer deployments for use cases such as analytics, file sharing, backup and archiving. Isilon's modular design dynamically scales from 16TB to 68PB by adding additional nodes in a single cluster. Through a set of data services and tight independent software vendor (ISV) integration, Isilon has been adding product features to expand product capabilities, from traditional home directories and active archiving to big data analytics and video surveillance workloads. Customers can choose a combination of several different platform models to match specific performance and capacity needs: high transactional, economical, high density or software-only. The latest OneFS release includes CloudPools, which allows customers to tier data to cloud storage, support for software-defined IsilonSD Edge and nondisruptive rolling upgrades. Isilon is leading the scale-out file system product market due to its stability, high data efficiency, management simplicity, scalability and data mobility, as well as a high level of customer satisfaction across many enterprise verticals.
ECS became available in June 2015 and now has 250-plus customers, many with over 1PB in their deployments. ECS has compatibility with HDFS, OpenStack Swift and Amazon S3, as well as APIs for Centera and Atmos. ECS can be deployed as a fully integrated appliance that scales well with a density of up to 3.8PB per single rack or as a software-only storage solution on certified commodity hardware of the customer's choice. ECS can be obtained as a subscription with a pay-per-use consumption model, and is now starting to gain wider customer and service provider adoption. ECS is positioned to address use cases that require distributed architecture, global namespace spanning across data centers with active/active access, geo-protection, multiprotocol access and metadata search.
- Isilon continues to grow enterprise adoption from traditional file system support to additional use cases, including video surveillance, big data analytics, backup and active archive.
- ECS has a proven track record of large production deployments across a variety of enterprises and is becoming an attractive price/performance platform for a variety of end users, including CSPs.
- End users looking to build a platform for unstructured data applications will find Dell EMC's broad portfolio of products, aggressive pricing, and new deployment and consumption models (such as software-only purchases, leasing, subscription and utility-based pricing) attractive.
- End users may feel the impact of sales or support changes as, postacquisition, Dell Technologies (which has acquired EMC) might pivot sales strategies, support offerings and R&D priorities for Dell EMC products.
- The current version of Isilon OneFS is not flash-optimized, and is not suited for small files and metadata-intensive workloads with low-latency requirements.
- Based on client inquiries, comparatively high ECS acquisition costs cause some Gartner clients to explore lower-cost object storage alternatives.
Hitachi Data Systems
Hitachi Data Systems (HDS), based in Santa Clara, CA, develops and markets the Hitachi Content Platform (HCP), an object storage product with enterprise file synchronization and sharing (EFSS) (HCP Anywhere) and a cloud storage gateway (Hitachi Data Ingestor [HDI]) components. HCP, first made available in 2006, is a mature product with deployments across a range of industries, including financial services, healthcare and public-sector enterprises. HCP's maturity and legacy as a compliance-focused archiving platform is supported by a broad range of ISV solutions, including content archiving, medical imaging, cloud storage gateways and enterprise content management (ECM) vendors.
In addition to using ISV applications, customers also build cloud-native applications using Amazon S3-compatibility supported by HCP. HCP Anywhere supports file synchronization clients on Windows, Mac, iOS, Android and Windows Phone. HDI can be placed at remote offices, and supports NFS and SMB protocols for access to data by applications and users while an HCP object store is managed centrally. In this regard, HDS is currently unique in that it both partners with vendors and develops its own EFSS and cloud storage gateway products. HCP offers a number of reporting capabilities that allow enterprises to separately charge back individual departments or applications for their specific usage. HCP integrates with both Active Directory and OpenStack Keystone for identity and token management services.
HCP supports a multitude of deployment options, but is most often delivered as a pre-configured appliance sold as a perpetual license or as a software appliance running as a VM in a VMware ESXi host. HCP can also be consumed with consumption-based pricing as a managed service, hosted private cloud, or public cloud service operated by HDS and other partners.
- HCP is a stable product supported by consistently positive marks by customers in terms of reliability.
- HCP has a large number of object storage customers across a diverse set of industries and workloads, offering enterprises a wide assortment of ISV applications and reference customers.
- Enterprises with requirements to provide mobile access to data and to support remote offices underpinned by an object storage platform can acquire these combined solutions from a single vendor, rather than deal with multiple vendors that 'meet in the market.'
- The average capacity of an HCP deployment and amount of data stored tend to be smaller than the Leaders in this Magic Quadrant. Gartner clients are not reflexively looking to HDS to solve the challenges related to managing petabyte-scale unstructured data with HCP.
- HDS does not serve the distributed file system segment of the market well and is behind the Leaders in terms of capabilities in this area.
- HCP doesn't have mind share among developers building modern, Mode 2 applications. In contrast, HCP is largely perceived as being ideal for Mode 1 workloads consisting of archiving and regulatory compliance.
China-based Huawei offers the OceanStor 9000 product, which has been available since 2013 and, according to the vendor, is now the fastest-growing product in Huawei's storage portfolio. Delivered as a hardware appliance, the OceanStor 9000 is particularly well-suited for the media and entertainment industries, and video surveillance. There are now more than 90 customers with greater than 1PB deployments.
Huawei OceanStor 9000 is a scale-out distributed file system product providing shared storage for unstructured data. OceanStor 9000 provides file services via NFS and CIFS shares, and object storage through Amazon S3 and OpenStack Swift interfaces. It features an all-active, share-nothing symmetric distributed architecture, where data and metadata are distributed to all the nodes to deliver performance and resilience at scale. The product scales from three to 288 nodes, with single file system support up to 100PB. Huawei is sold as a hardware appliance only and has several different hardware node options, each optimized for small capacity, high IO/s; large capacity and high bandwidth; or video surveillance, and archiving, and supports intermixed node deployment.
In September 2015, OceanStor 9000 added object (S3 and Swift) and object deduplication support under the same unified storage management umbrella. Despite this, OceanStor 9000 is still primarily being deployed for file system workloads requiring high bandwidth and scalability.
- The OceanStor 9000's highlights are high scalability and resilience, supporting up to 100PB of storage on up to 288 nodes based on a symmetrical distributed architecture.
- The OceanStor 9000 is particularly suitable for enterprises in the AsiaPac region, as Huawei has expanded its presence and witnessed a triple-digit increase in its year-over-year revenue.
- The OceanStor 9000 is well-suited for video surveillance, media workflow and media asset management use cases.
- Huawei's long-term storage product roadmap remains uncertain for enterprise IT buyers. This has included its recent decision to discontinue its OceanStor UDS, Huawei's dedicated object storage product line that has struggled to gain traction.
- Huawei's service, support and reseller network for storage products continues to be weak in the North American market, where it still struggles to grow its market share due to geopolitical and brand perception issues.
- The OceanStor 9000 object protocol is less mature than file, as the majority of its existing end users are only utilizing the file protocol portion of the unified storage product.
IBM has two products included in this research - IBM Spectrum Scale and IBM Cloud Object Storage. IBM Spectrum Scale is based on General Parallel File System (GPFS), a distributed file system built by IBM researchers in the late 1990s. IBM Spectrum Scale is highly scalable and delivers competitive concurrent throughput performance due to its decentralized architecture and efficient data placement across the nodes. The product also supports information life cycle management by supporting IBM Spectrum Archive (tape) and tiering to cloud storage. While GPFS has long been sold as an integrated hardware platform, the recent focus on software-defined deployment models - both on-premises and in the IBM Cloud - has the potential to increase the attractiveness of the product to prospective buyers.
IBM's acquisition of Cleversafe in 4Q15 gave IBM a competitive, mature, proven and scalable product in the fast-growing object storage market. The product has been rebranded as IBM Cloud Object Storage. Cleversafe's market differentiation is protected through a robust patent portfolio, with the vendor having been granted more than 400 patents. Cleversafe has several petabyte-scale customers, with a few exceeding 100PB in installed capacity. By implementing a distributed erasure coding (information dispersal) algorithm, the product enables HA at lower-capacity overhead, with the ability to tolerate site failures. IBM Cloud Object Storage supports the S3 and Swift APIs on the front end for access. IBM Spectrum Scale can now tier to IBM Cloud Object Storage, with further integration being one of the key roadmap goals for the product teams.
- IBM Spectrum Scale and IBM Cloud Object Storage are mature products with flexible deployment options and hundreds of PB-scale customers.
- IBM Cloud Object Storage delivers a differentiated way to secure data at rest. It's all or nothing (AONT) encryption is integrated with its information dispersal algorithm to lower the potential for data breaches,
- IBM Spectrum Scale delivers competitive throughput performance for both small and large files; provides broad support for interfaces, including NFS, SMB, REST (S3, Swift) and HDFS; and can tier data to Amazon S3 and the IBM Cloud Object Storage.
- The IBM Cloud Object Storage product lacks native support for file access and needs thirdparty tools to enable it.
- Based on Gartner client feedback, IBM Spectrum Scale's licensing is difficult to understand with several editions of the software that are licensed on a per-socket basis with a mix of different license types (server, client and File Placement Optimizer [FPO] licenses), based on the role performed by the node.
- Cleversafe was acquired by the SoftLayer division of IBM. This may result in future product development priorities for IBM Cloud Object Storage to be more closely aligned to the needs of SoftLayer than that of enterprise IT.
Based in Sunnyvale, CA, NetApp remains one of the leaders in ECB block and file storage. NetApp's StorageGRID Webscale is an object storage product providing API-accessible, distributed storage with the ability to scale across geographies. StorageGRID Webscale supports the major object protocols (and offers NAS access) on dedicated hardware or as VMs in either VMware or OpenStack environments. The origin of NetApp StorageGRID Webscale comes from the acquisition of Bycast in 2010, with a strong footprint in healthcare records storage and management.
StorageGRID Webscale 10.2 is a distributed content repository using object storage software with metadata stored in a distributed NoSQL database with distributed access. Up to 70PB of storage and 100 billion objects across up to 16 geodispersed sites in a single namespace are supported. StorageGRID Webscale also supports object archiving to tape and/or any S3-complaint cloud storage. The NetApp AltaVault cloud gateway product can be integrated with StorageGRID Webscale as a target for backup and archive. StorageGRID Webscale supports multiple deployment models, including software-only deployments and NetApp-engineered appliances, with an ability to mix and match both in the same cluster.
NetApp released StorageGRID Webscale two years ago, and it has been rewritten to correct the architectural scalability limitations of its preceding generations. The latest improvements include S3 API enhancements; support for CIFS, NFS and Swift APIs; geodistributed hierarchical erasure coding; multitenant identity federation; and policy-driven cloud tiering.
- NetApp has a mature service and support infrastructure, and a presence in 100 countries worldwide.
- The new StorageGRID Webscale product has been redesigned to improve metadata handling, such as scalability and resilience.
- StorageGRID Webscale has unique identity and access management (IAM) capabilities that mimic those of AWS, resulting in a seamless transition between public cloud and on-premises object storage environments.
- Substantial parts of StorageGRID Webscale have been rewritten and are currently lacking largescale reference customers.
- StorageGRID Webscale lacks integration with the NetApp Ontap product line to provide enterprises with a unified distributed file and object storage experience.
- StorageGRID Webscale lacks a strategic focus within NetApp compared to its other products with more market penetration, which impacts the resources allocated to develop StorageGrid and bring it to market.
Panasas, based in Sunnyvale, CA, is a privately-held storage company that has primarily focused on the HPC storage market. Its flagship product is the ActiveStor appliance, which is based on its PanFS distributed file system. The vendor's focus has been on building high-performance storage products with high scalability, availability and ease of use. Due to its flash-enabled architecture and through its proprietary protocol that enables direct, parallel access by clients, ActiveStor delivers robust performance with support for both IB and 10GbE. Its RAID-6+ data protection scheme offers triple parity protection with rebuild performance that can scale linearly. The product is sold based on raw capacity with all-inclusive licensing, while the actual usable capacity varies based on the end user's choice of data protection level.
The key target industries where Panasas has had market traction include energy, public sector, life sciences, manufacturing, media and entertainment, and research. In 2015, Panasas made a strategic shift to focus more on the commercial market due to stagnant revenue in the HPC space and declining gross margins. In order to serve enterprise customers, the vendor has been focusing on driving channel partner growth with new partners such as Ingram Micro.
- Panasas is an established leader in the high-performance storage space, providing integrated hardware and software support with deep workflow integration with applications in industries such as life sciences, manufacturing, energy and the public sector.
- ActiveStor supports multiprotocol access including NFS and SMB, as well as the vendor's proprietary protocol, Direct Flow, which enables high-performance parallel data access from clients to the storage system.
- Due to its object-based file system architecture and distributed approach to erasure codes, Panasas is able to protect data at a file level, resulting in lower-capacity overheads and faster rebuilds.
- Panasas' historical focus on the lower-margin HPC market can impact its ability to scale investment in the highly competitive commercial scale-out NAS market.
- Panasas delivers limited data services and has weak integration with horizontal ISVs in backup/ recovery and archiving space.
- The ActiveStor product neither supports RESTful APIs on the front end nor does it support any kind of integrated cloud tiering.
Based in Raleigh, NC, Red Hat has two open-source products that have been included in this research, both of which have come through acquisitions. Red Hat acquired Inktank in 2014, a provider of maintenance and support for the popular open-source Ceph project. Red Hat Ceph Storage, as it is now called, delivers unified block, file and object access in a single system with an ability to scale out to PBs of data, all on commodity hardware. Ceph's architecture is underpinned by RADOS, a distributed object storage service, which uses a unique algorithm to handle data placement that balances availability and performance. The Ceph project is supported by a vibrant open-source ecosystem with aggressive software release cycles and robust community support. OpenStack private cloud implementations have been the primary focus of the Ceph community, where Ceph continues to enjoy wide popularity and broad support from distribution vendors for OpenStack VM storage. In addition, other common use cases include content distribution, backup and archiving target. Red Hat Ceph is licensed based on capacity with subscription-based pricing that is competitive when compared to the other proprietary products in this space.
Red Hat acquired Gluster in 2011, and the product is now called Red Hat Gluster Storage. It is a scale-out, multiprotocol (NFS, RESTful APIs, SMB), open-source storage software solution, with petabyte-scale capacity and competent data protection (snapshots and replication) capabilities. Following Red Hat's acquisition of Inktank, the Red Hat Gluster Storage product now primarily targets file protocol- based use cases. Through a regular cadence of software releases, Red Hat has kept the product competitive with last year's software release, including important capabilities such as erasure coding, automated tiering with support for flash and easier snapshot scheduling. Red Hat Gluster Storage has a node-based licensing scheme. As a first integration step, Red Hat recently released a common administration platform with Red Hat Ceph Storage 2, which is intended to manage both products in the future.
- Red Hat has strong enterprise pedigree with a proven commercial model for open-source software with a differentiated storage portfolio that is focused on bimodal requirements of enterprise IT.
- The Ceph project has an active community of code committers and a growing ecosystem of OEM partners, such as Cisco, Dell, Fujitsu, SanDisk and Supermicro, that have built either reference architectures or appliances.
- Red Hat Gluster Storage offers customers flexible deployment options - bare-metal install, virtual storage appliance on KVM, VMware hypervisors, AWS and Microsoft Azure public clouds, and as a Docker image. It can also be deployed as a hyperconverged appliance through third-party server OEMs.
- Ceph deployments often require a major effort from enterprise users, calling for careful instrumentation to optimize and fine-tune cluster performance and reliability for chosen hardware.
- The Ceph Filesystem (Ceph FS) isn't mature enough to be deployed for production use cases in the enterprise due to limited data protection features. It is in a tech preview phase and only recommended for OpenStack file-as-a-service workloads.
- Gartner client inquiries reveal confusion over the growing product overlap between Gluster and Ceph, further heightened by the recent announcement of Ceph FS.
Scality is a venture-backed, privately-held company that develops object and scale-out file storage software, with R&D in France and sales and marketing based in Silicon Valley. First released in 2010, Scality Ring is scale-out, peer-to-peer distributed shared-nothing software that can be deployed on any x86 commodity hardware; HPE and Dell are formal resellers of this technology, and have predefined Scality SKUs. At the end of 2015, HPE had also invested in Scality. Ring has more than 100 large customers, including service providers, public sector, life sciences, financial services, research institutions and large enterprises. IT leaders have implemented Ring for email hosting, content distribution, active archive and private storage cloud.
Ring supports configurable erasure coding and replication in a peer-to-peer architecture that provides efficiency at scale, with no single point of failure. Ring is based on a standard Linux distribution nonmodified kernel, and supports standard protocols for objects (S3, CDMI, file [NFS, SMB, FUSE]) and OpenStack (Swift, Glance and Cinder). Unlike many competitive products, Ring supports native file system access to Ring storage through the file connector servers and the integrated virtual file system that provides file storage services without the need for external file gateways, as is commonly required by other object storage products.
Ring is best-suited for large multi-petabyte deployments with geographically distributed sites, with some reference customers reaching 30PB deployments. Throughout its existence, Scality has been focusing on improving product performance for challenging workloads like small files and random I/O. In a recent release, Ring enhanced its file system support, including automatic failover, and improved parallel and multiuser write performance to the same file system directories. In addition, Scality S3 API now supports integration with Microsoft Active Directory and with AWS IAM, providing multi-tenancy for enterprise user groups.
- Scality has strong channel partnerships and OEM relationships with Dell and HPE.
- Ring provides a native unified (file and object) platform for enterprise customers.
- Ring is designed to provide good performance for small files, which allows it to target replacing traditional NAS for unstructured and semi-structured data.
- Customer sales and support experience in North America has been mixed due to high sales turnover and significant changes to Scality's workforce in 2015.
- Ring is not designed as a replacement for general-purpose NAS solutions, as it is not optimized for heavy metadata and random access workloads, and does not have all the features of enterprise NAS (snapshots, NFS v.4 support).
- Scality is the largest independent distributed file system and object storage vendor that does not sell its solution as a pre-integrated hardware appliance; as a result, it is a potential acquisition target for a larger vendor seeking to add products and expertise focused on an emerging market.
SUSE, based in Nurnberg, Germany, is a contributor to Ceph, an open-source storage software project, which SUSE Enterprise Storage (SES) is based on and which was first made available in 2015. SUSE is focused on adding installation and management capabilities to its contributions to Ceph. Ceph is mostly used as block storage, with the primary use cases being databases and VM storage, but Ceph's block storage interface is based on an object storage foundation. As such, Ceph is sometimes used as a stand-alone object storage platform as well, since it supports the Amazon S3 and OpenStack Swift protocols that are commonly used in object storage deployments. Most customers that use Ceph primarily for its object storage capabilities are comfortable deploying open-source software on commodity hardware themselves and managing the complexities that this can entail. Often, customers rely on the Ceph community for support, but occasionally customers start down this route and then seek support from an organization such as SUSE.
The base configuration of SES includes four storage nodes and six infrastructure nodes that include monitoring, management, object storage, iSCSI and metadata services. Customers can expand on this base configuration node by node for a nominal price. SUSE has a partnership with it-novum to jointly provide Ceph management capabilities using the openATTIC open-source project. SES is available on Apollo and Proliant servers through an OEM relationship with HPE and sales partnerships with Cisco and Lenovo.
- SES is aggressively priced, based on nodes rather than capacity, and, as a result, customers can deploy cost-efficient object storage using commodity servers with densely configured disk storage.
- SUSE is focused on making OpenStack more enterprise-consumable through efforts such as improving backup and recovery with Ceph.
- SES may appeal to existing customers of SUSE Linux Enterprise (SLES), a frequently deployed OS particularly for Mode 1 applications.
- The majority of SES deployments are in Europe, whereas the product has negligible adoption in North America, reflecting SUSE's recent entry in the market for enterprise storage and a lack of credibility in the space.
- Few Gartner clients are considering SES for its object storage capabilities as most of SUSE's Ceph customers use the product primarily as block storage.
- SUSE has overly ambitious plans with a forthcoming release that includes support for Ceph FS, which is largely considered to not be enterprise-ready by other upstream contributors to Ceph.
SwiftStack, based in San Francisco, CA, is the main contributor to Swift, an open-source object storage project in the OpenStack suite of cloud infrastructure software. SwiftStack, in its commercial role, develops a product that allows enterprises to better deploy and manage a SwiftStack cluster, which first started shipping in 2013. Although the foundation of SwiftStack is based on OpenStack Swift, the vendor has also adopted the Amazon S3 API because of the sheer size of the S3 ecosystem of integrated tools and the market opportunity, compared to Swift.
SwiftStack supports backup and archiving to its platform using Commvault and Veritas' NetBackup. SwiftStack also offers a virtual drive that provides desktop users with drive letter access to the object storage repository. Although SwiftStack works with other cloud storage gateway vendors such as Ctera, it also develops its own SMB- and NFS-based cloud storage gateway that has the benefit of access interoperability between the file and object storage protocols. This allows, for example, videos or other rich media to be created using NFS, but read using the Amazon S3 API.
The profile of an enterprise customer that deploys SwiftStack is different from an enterprise that deploys competing Ceph offerings as a stand-alone object storage platform. Although both products have substantial open-source foundations and are commonly deployed with OpenStack, the profile of a SwiftStack customer is more mainstream than that of Ceph when used as standalone object stores. When customers evaluate SwiftStack, they are typically focused on attributes such as TCO, company viability and support. The open-source nature of Swift is often of secondary importance when customers are considering SwiftStack. SwiftStack has a partnership with Cisco, whereby Cisco resells SwiftStack with its UCS servers and as part of its Metapod offering that is deployed on-premises, but operated remotely by Cisco.
- SwiftStack has developed significant capabilities to reduce the costs of implementing and operating an OpenStack Swift cluster.
- Customers consistently rate SwiftStack highly for reliability, overall relationship with the vendor and satisfaction from the presales phase to product deployment.
- Despite being a comparably smaller company and salesforce, enterprises are beginning to consider SwiftStack alongside the other, more established vendors with much larger sales operations.
- Swift requires significant tuning to optimize performance for particular hardware and workloads, and this often limits the appeal of SwiftStack to enterprise buyers.
- SwiftStack has focused much of its efforts on the Amazon S3 API due to the sheer size of the S3 ecosystem and the market opportunity, leaving its commitment to OpenStack Swift API origins in question.
- The vendor has wide support for server and storage hardware that may result in support issues for configurations that are not commonly deployed or tested.
Vendors Added and Dropped
We review and adjust our inclusion criteria for Magic Quadrants as markets change. Because of these adjustments, the mix of vendors in any Magic Quadrant may change over time. A vendor's appearance in a Magic Quadrant one year and not the next does not necessarily indicate that we have changed our opinion of that vendor. It may be a reflection of a change in the market and, therefore, changed evaluation criteria, or of a change of focus by that vendor.
Inclusion and Exclusion Criteria
To qualify for inclusion in this Magic Quadrant, vendors must meet all of the following requirements:
- Revenue must be above $10 million per year for the distributed file system and/or object storage product between 1 May 2015 and 30 April 2016, or there should be at least 50 production customers with 300TB or more for utilizing distributed file or object storage protocols.
- The product must be installed in at least two major geographies (among North America, EMEA, the AsiaPac region and South America).
- The vendor should own the storage software intellectual property and be a primary software developer for the storage product, with an exclusion for open-source software.
- The product must be sold as either an appliance or a software-based solution.
- The product must be available for purchase as a stand-alone storage product, and not as an integrated or converged system with a compute and hypervisor bundle.
The product must have:
- File and/or object access to the common namespace/file system.
- A shared nothing architecture where data is replicated or erasure coded over the network across multiple nodes in the cluster. The product must also have the ability to handle disk, enclosure or node failures in a graceful manner, without affecting availability.
- A single file system capable of expanding beyond 300TB.
- A global namespace capable of 1PB expansion.
- A cluster that spans more than three nodes.
- Support for horizontal scaling of capacity and throughput in a cluster mode or in independent node additions with a global namespace/file system.
Vendors to Watch
The following vendors did not meet the criteria for inclusion in the Magic Quadrant, but may be worthy of consideration as they are starting to get more traction in the unstructured storage market:
- Formation Data Systems