Nexustorage aims at file and block access to object storage

Cloud and object storage bring numerous benefits, such as massive scalability, flexibility and reduced cost of storage. But many – in fact, most – applications are just not written to take advantage of object storage.

New Zealand-based Nexustorage aims to remove that gap and allow customers to gain file and block (NFS and iSCSI) access to data that is predominantly held on in-house or cloud-resident object storage, according to a recent presentation on the IT Press Tour virtual edition.

That’s useful for applications and database workloads that need file and block access to data, but in which much of the data held can be stored longer-term on object storage, in-house or in the cloud.

Key claimed benefits of the software-defined storage products are: reducing the cost of primary storage, with data being tiered off to cheaper layers; reduction in network traffic; and cutting costs and transactions to the cloud, with only changed data being moved. Migration between tiers can be scheduled for quiet periods so the working day is not disrupted.

Above all, perhaps, is simply the ability to access large, relatively cheap and scalable object storage and make use of it in traditional file/block-based applications.

Core to what Nexustorage does is a tiering system courtesy of its Nexfs file system that presents a single pool of storage, but in which files are broken down into chunks (1MB by default, but up to 8MB) and tiered to SSD and/or SATA spinning disk, with a bulk backing store of S3 object storage.

So far the company is yet to release a commercial product, but a community edition is planned for release this month. Eventually, customers will be able to access all data in a cluster from anywhere, but currently, you can only get to it from the server that created it.

Nexustorage founder Glen Olsen said: “You can’t do block-style workloads on object. For example, you can’t use object to run the C drive of a running Windows server, or your database. That’s what we’ve changed – to work out how to integrate block workloads with object storage.”

Olsen is referring to the fundamentals of data storage, which was for decades based on file system access to drives with the familiar letter-based drive format and tree-like directories containing files.

File access storage is quite obviously based on this, but so is block storage which accesses parts of files within that structure. However, object storage is quite different and relies on a flat structure of objects with their own unique identifier, to which some kind of index or directory is held.

File and block access via Posix is the way applications have been written forever, but they can’t access object storage in the same way – with file locking and other protections, for example.

Meanwhile, object storage is increasing in use, not least because it underpins data storage in the cloud, and while applications are being written to take advantage of it, the vast bulk still require Posix access.

That need is what Nexustorage aims to address with its Intelligent Data Placement. All data is broken down into sub-file chunks. Everything is held in the object tier – and optionally accessible via an open source S3 proxy – while more frequently accessed, written or read data chunks can be retained on SSD or SATA tiers for more rapid access.

“Other solutions are outdated, and move the complete file between local and cloud storage, leaving a stub, for example,” said Olsen.

Nexustorage looks at files and makes decisions about which tier data should go on, according to an input/output (I/O) access profile that characterises the likelihood of it being recalled. The company claims it can do away with a separate index or manifest of file locations with a feature called Nexassert.

“It knows where data is without recording where it is,” said Olsen, commenting on the patent-pending Nexassert.

Containers are not supported, although this is planned for a future release.