EvgV - stock.adobe.com
Scality Ring XP to offer object storage at scale with rapid access
Lightweight S3 API trades storage features for microsecond access. Scality pairs it with all-NVMe flash in Ring XP, aimed at AI use cases for small object data in PB-scale datasets
Object storage specialist Scality has launched Ring XP, a tuned-up version of its existing Ring platform aimed at artificial intelligence (AI) data lake use cases that promises microsecond (µs) storage access times.
That tune-up centres on the use of a stripped-down version of S3 object storage application programming interfaces (APIs) – that lack performance-sapping functionality – and pairing the platform with all-NVMe flash hardware.
According to Scality chief marketing officer (CMO) Paul Speciale, the launch is driven by customer demand for PB-scale data lake aggregation with rapid input/output performance for things like data cleansing, security, immutable backups and extreme metadata handling. “Ring XP is a special configuration of Ring for small object data found in AI processing,” he said.
Small object data here comprises, for example, fragments of imagery, character and numeric strings in logs. In short, very small chunks from very large, PB-scale, datasets.
Speciale said that in this arena, Scality XP can compete with high-performance storage from the likes of IBM, Pure Storage and Vast Data.
Core to the performance promised by Ring XP is a stripped-down S3 connector and all-NVMe flash storage.
At one end of the scale is full-fat S3 API connectivity. This comes with storage functionality such as versioning, replication and other management features. This is what you get with Scality Ring (i.e. not XP) and Amazon S3, where access times are measured in the 10ms–50ms range.
Read more on object storage
- File, block and object: Storage fundamentals in the cloud era. We look at the three basic ways that storage accesses data – via file, block and object – as well as the ways in which the rise of the cloud and distributed systems have brought changes to them.
- Is object storage good for databases? We look at object storage vs block access SAN storage and ask if object storage can be used for database workloads, or is it just good for bulk storage of analytics datasets?
It is possible, however, to access versions of S3 that dispense with that advanced functionality and which provide higher levels of performance. AWS, for example, offers S3 Express Zone One, in which reduced features and a single availability zone provide latency of between 1ms and 9ms.
Ring XP goes further, and promises less than 500µs for gets and less than 700µs for writes. “The Amazon S3 stack is very rich but introduces 10s of milliseconds of latency,” said Speciale. “Amazon recognises that and provides faster options. We can give even faster ones, although you can’t do it without the right hardware. We don’t see another storage provider publish latency figures that low. That’s what we base our claim to be the fastest on.”
Existing Ring customers can turn on Ring XP functionality by paying the licence cost – if they have the correct hardware available.
Scality’s two products are Artesca and Ring, which aim at opposite ends of the same market – namely, enterprise and small and medium-sized business customers that want object storage, but also want to keep applications on-premise.
Scality Ring is aimed at environments of at least 5PB. Artesca focuses on object storage for backup applications.
Examples of Ring users provided by Scality include a US banking customer that uses Ring as an aggregation for all of its data subject to fraud detection via Splunk, with capacity into tens of PBs. European air travel pricing optimisation platform Amadeus uses Scality for its Splunk-based work and data turnover of 1PB per day.
Elsewhere, French genomics research firm SeqOIA uses Ring as its bulk store, but also, according to Speciale, it’s fast enough to carry out some performance-hungry operations. SeqOIA retains a Vast Data deployment that can handle extreme high-performance requirements, but according to the CMO, it hasn’t grown from its initial capacity because Scality Ring can handle the scale of data and many operational workloads.