In October, Germany-based PoINT will launch an object storage archiving product based on disk and tape media.
“Our idea is to offer on-site exactly the same thing as AWS [Amazon Web Services] offers online with S3 and S3 Glacier,” said Thomas Thalmann, PoINT CEO and co-founder.
Archival Gateway – Unified Object Storage uses high-capacity, low-cost disk to store cold data. That is, data the enterprise will only access rarely.
Like S3 Glacier, tape is also used. It’s slower than disk but also cheaper.
PoINT said it can restore files in milliseconds from disk and in minutes from tape. To determine which files go to which media, PoINT’s software analyses something like 50-plus parameters using pre-defined rules.
For example, a user may want files of a certain type to migrate from disk to tape after 30 days of inactivity, or another type of file to be deleted from the archive after a set expiry date. Or, they may want files to be copied to both media but only to stay on disk for as long as their analytics work is carried out.
Compared with archiving services in the public cloud, PoINT’s service guarantees that data doesn’t go beyond the enterprise firewall – and the enterprise owns the disk and tape equipment rather than having to continually pay fees to the cloud provider to keep the data in its infrastructure.
230GBps with just tape
Archival Gateway – Unified Object Storage is the first product of this type to write object storage buckets directly to tape.
The majority of other object storage products capable of archiving contents to tape do it by taking a backup and copying that in file format to a tape library. That makes the process more lengthy and means restored data will not be in object format. To retrieve object metadata and provide S3 access, it must be restored to the S3 cluster. That means several stages in an operation that are often dictated by conditions of urgency.
“With Archival Gateway you can build an S3 cluster of eight tape libraries, each with eight readers working in parallel,” said Thalmann. “As we do with erasure coding at the level of tapes, we parallelise access. This solution, even without hard disks, can achieve throughput of 230GBps.”
Archival Gateway – Unified Object Storage is the second version of a product launched in Germany in 2018 that didn’t have HDDs, only tape.
“For our German customers, the novelty that we’re announcing is the arrival of tiering, with less latency on the disks and around 50 rules to pass from one media to the other,” said Thalman.
For customers elsewhere, Archival Gateway is a completely new S3 storage product that – owing to the large number of tape cartridges that can be accommodated – can manage up to 50 billion objects per bucket.
Archival Gateway comprises gateway nodes that present S3 object storage on the network. Each of these nodes accesses a maximum of eight read heads. The gateways are backed up by so-called database nodes, with a maximum of four per cluster. These index contents help direct the read heads and run the admin console for the offering.
A descendant of Philips and Sony
Behind the news, PoINT has some illustrious history in the archiving sphere. Its product, Archival Gateway, is a descendant of another archiving offering – Everspan Gateway – that the company developed for Sony in 2016. Back then, this was not a cluster of tape readers but comprised 64 BluRay optical disk writers that gave a performance of 18GBps and indexed 181PB of data.
PoINT made its name originally during the mid-90s boom in optical disk writers, of which one – CDWrite – achieved global success on Windows. Before founding PoINT in 1994, Thalmann and his team worked at Philips, where they developed the fundamentals of CD-R, the first writable CD-ROM.
Besides Archival Gateway, since 2007, PoINT has also sold another storage product: PoINT Storage Manager. This identifies production data on the network, and pinponts which is in use and which is cold. Dependent on this analysis it moves data to different tiers of storage, and at the end of its lifecycle, to disk or tape archives.
PoINT Storage Manager claims two unique attributes. The first is that it leaves symbolic links on the production array to data that has been moved. That means users still see files where they expect them, even though they have been tiered off elsewhere.
“These aren’t simple symbolic links in the sense of Windows or Linux,” said Thalmann. “These are links that use proprietary APIs from a large number of NAS arrays that include NetApp and Dell EMC.
“This system allows us, when a user wants to reopen a file, to gradually bring the required blocks closer. Our archives are immutable, to resist ransomware. And if the user modifies the contents, we generate a new archive that uses all the same links.”
The second feature is the format of the archive, which is a disk image with a universal disk format file system inside.
“To schematise, most archive solutions pack your data into a zip file or equivalent,” he said. “The problem with that is you only have raw files within. If you pack things into a disk image, you also archive with them their date of creation, modification, reads, as well as their author and access rights.”