Photobank kiev - Fotolia
Spectra Logic has launched a new bulk storage array, the Verde DPE, which it claims can provide storage for as little as $0.09 per GB.
DPE – which stands for digital preservation: enterprise – is aimed at customers that want to store large files and unstructured data in disk archive.
The product – effectively archive-class network-attached storage (NAS) – uses the relatively slow access 8TB SMR (shingled magnetic recording) drives from Seagate and is aimed at the tier of storage below existing Spectra Logic Verde products which use SAS drives for more rapid access to data.
With the 8TB drives, Spectra Logic can pack up to 7.4PB in a rack with around 200TB of capacity per 1U of rack space and throughput of 1 GBps.
Spectra Logic chief technology officer Matt Starr said customers could expect storage cost per GB of $0.10 at 1PB of capacity and $0.09 at 7.4PB.
Data protection is potentially an issue with such high capacity drives, especially during rebuilds following disk failures. This is addressed by the ZFS-based Raid Z3. Verde DPE comes in “bands” comprising 25 drives, of which 20 are reserved for data, three for parity data and two as spares. Data is striped across the 20 drives and data rebuilt using the parity data on the other three drives. Spectra Logic claims that produces likely data loss of once in one or two million years.
Verde DPE is based on the ZFS file system laid on a BSD-based operating system.
In SMR drives, data tracks do not sit next to each other as in standard hard disk drives (HDDs), but are overlaid to increase capacity. This, however, adds to write times so metadata that must be accessed more frequently is stored on areas of the drive with tracks arranged in the standard way. The Verde controller also arranges random writes into sequential patterns to cut down on disk head movement.
Access protocols supported are NFS and CIFS. These are NAS protocols and Verde DPE is being marketed as a NAS replacement.
Did Spectra Logic consider object storage for the use cases it has in mind? Increasingly, object storage is emerging for the storage of very large amounts of unstructured data, which is usually unstructured. Its method of storage – via a flat structure with unique identifiers – is widely regarded as superior to the tree-like file system in NAS, which can become unwieldy at high volumes.
Starr said: “This isn’t object storage, but that would be a good space for this product. That’s not to say it won’t be integrated with Black Pearl or object storage. We will see products similar to this, but in the object storage space in the future.”
Read more about archive storage
- Cloud archiving is shaping up as one of the “killer apps” for cloud storage, but don’t be tempted by capacity alone.
- Latency in object stores that stems from a large amount of metadata means the technology is better suited to non-transactional data.