Scality Updates Ring Organic Storage

Scality unveils a new version of its RING Organic Storage, which addresses all needs of file storage at scale.

While file storage has traditionally been deployed on dual-controller NAS architecture, the need for scale, which results from big data applications and the move to cloud infrastructures, requires a more scalable design.

With RING Organic Storage 4.0, Scality delivers a solution that handles both the primary storage and the long-term preservation of files at petabyte scale, with a density of up to 1.8PB of protected data per rack. RING is a effective than traditional NAS or scale-out NAS because it leverages commodity hardware and reduces cost of operations.

RING Organic Storage 4.0 comes
with 4 new essential features:

• A complete scale-out file system
• Advanced Resilience Configuration (ARC) for efficient long-term storage
• A performance unsurpassed by any object based storage to date
• A new management system to simplify the management of hundreds of servers

And the 4.0 release continues to leverage properties of Scality’s design such as parallelism, self-healing, hardware agnostic organic upgrades and multiple datacenter replication for disaster recovery.

Industry analysts agree that the storage of files or unstructured data now represents the bulk of storage required by service providers and large enterprises. At the current pace, unstructured data will soon reach 80% of all storage requirements. Traditional NAS storage has had to resort to overly complex architectures to handle this data volume, which has led to unsustainable system sprawl.

"Scality RING Organic Storage is a revolutionary step in the storage of files at scale," said Jerome Lecat, Scality CEO. "For the first time, storage architects can select a solution that delivers performance for hot data, a cost of ownership comparable with webscale infrastructures, and simplicity of management, all at the same time."

"Scality RING has seen an accelerating rate of adoption over the last few quarters particularly for primary storage applications. We expect the new RING Organic Storage 4.0 to appeal to our historical customers, as well as to web hosting service providers and large enterprises as they require more scalable file storage," added Lecat.

Scale-out file system
RING 4.0 introduces standard file-based access thanks to a scale-out file system. Unlike other storage vendors who deliver multiple interfaces by placing a gateway in front of their underlying storage technology, Scality is delivering file access with its own parallel design. As a result, the Scality file system benefits from the same performance, fault tolerance, self-healing, and organic growth characteristics intrinsic to its pioneering RING platform. Particularly, the ability to deploy a large number of connectors with shared metadata on a global namespace assures the delivery of virtually unlimited IOPS for file applications.

"By integrating as a userland agent within the Linux FUSE framework, our scale-out file system provides 100% POSIX compliancy. Additionally, by using a low level FUSE interface, we’re able to serve highly parallel workloads very efficiently as concurrent operations are spread out among all RING nodes. Sparse files and random writes at any offset are made possible thanks to MESA, the RING’s unique distributed transactional database," said Giorgio Regni, Scality’s CTO.

Advanced Resilience Configuration
ARC was unveiled as part of the RING 4.0 release. It represents a new way to protect data against failures. This feature is based on erasure code technology, which has been used in the telecommunications industry for some years. At very large scale, RAID-5 and-6 present vulnerabilities in terms of risk exposure and cost, which makes them unviable for such use cases.

Scality introduces a solution in which the original data is preserved and coupled with multiple checksums from which the data can be derived. The benefit is that original data can be read directly, rather than calculated from the checksums. This is a fast operation without any ‘penalty-on-read’ typically associated with additional cpu cycles for data reconstruction.

Scality chooses a (16,4) combination by default, where 16 represents the number of original data fragments and 4 the number of tolerated failures. In this configuration Scality ARC accommodates 4 simultaneous failures, whether they result from disks, servers, networks, racks or sites. In terms of storage consumption, ARC (16,4) adds only 25% more storage than raw data globally, a reduction from 2 or 3 copies in a replication environment.

ARC also delivers at least 99.9999% of durability for local configuration and reaches 99.99999999999% when ARC is replicated to another data center.

"At petabyte scale, it is cost prohibitive to deploy a replication-based infrastructure. Scality ARC represents the perfect solution for such environments. Delivering high performance with direct I/O operations, ARC provides high levels of protection with a substantial economy," confirms Brad King, director of Customer Architecture at Scality.

ARC is also the solution for space-constrained environments. With hardware from manufacturers, holding 72 HDD of 3TB and a motherboard with 4 rack units (4U), Scality is able to deliver over 1.8PB of protected storage within one rack.

Performance
Scality’s reference architecture is based on a two-tier architecture, with a first tier handling active data and file-system metadata and a second tier delivering web scale cost effectiveness for long term data storage. This two-tier architecture is transparent to client applications, which only see a virtually unlimited storage pool.

In a report published today, ESG Lab has evaluated the performance of Scality’s first tier when deployed with SSD and found that it delivers no more than 7ms latency for complete read or write operations on 4KB objects, and a throughput comparable with high performance computing systems.

In the conclusion of its report, ESG Lab writes: "Until now, object storage could deliver the scalability, but not the performance. Standard response times for object-based storage are in the hundreds of milliseconds to full seconds, much too slow for today’s cloud-based services. In comparison, ESG testing demonstrated that Scality object storage on Intel Xeon servers equipped with Intel Enterprise SSDs and low-latency 10GbE network environment delivers 4-10 millisecond performance – 10 times faster than other scale-out systems. The Scality solution delivers scalable object storage in which performance is never a barrier. Users can solve any performance problem with this architecture just by selecting the right number and types of server nodes and disk."

Daniel Binsfeld, VP Customer Services of Scality notes: "In practice, performance is not an issue for our customers, because we deliver much more than they typically require for their file-based use cases. Our two tier architecture delivers an average of 40ms for 100 KB object writes and reads when deployed solely on HDD, and under 20ms when deployed on a mix of SSD and HDD."

New Supervisor features
Leveraging the experience gathered from existing customer deployments, Scality has redesigned its management system, known as the Supervisor. This management platform, which has its own web GUI, can also be integrated in existing customer management systems using a REST interface, SNMP or command line functions and scripting. The new Supervisor has simplified the management of a large number (thousands) of servers.

Furthermore, the new management platform keeps track of more indicators on the health and operations of each storage node, simplifying performance tuning, troubleshooting and capacity planning operations.

Jérôme Lecat concludes: "Scality RING 4.0 is a major evolution for Scality but also represents a redefinition of object storage. We fully expect this new development to influence the way that end users view the use case for object storage. It further demonstrates our capacity to develop truly innovative technologies, perfectly aligned to the growing market for file-based storage at large scale."

Scality RING version 4.0 has been deployed in beta since January 2012, and is available. It continues to be priced per capacity on a pay-as-you-go basis.