Seagate and pals form key-value hard-drive Kinetic gang
Kinetic Open Storage Project is not the OpenKinect Project
Three disk-drive giants and eight other IT suppliers have set up an alliance to promote directly-addressed object-storage – hard drives that store data in the form of keys and values, in other words.
The eleven have formed the Kinetic Open Storage Project (KOSP), which has joined the Linux Foundation as a collaborative project. The eleven suppliers are:
- Seagate, which first invented the idea of Ethernet-accessed key-value store drives with its Kinetic disk drive.
- Toshiba, with its own Ethernet key-value store drives.
- WD, whose HGST subsidiary has its own directly-addressed object storage disk drive technology.
- Cleversafe, an object storage software supplier.
- Dell, which is setting up a reselling deal with Scality.
- NetApp, which has its StorageGRID object storage product.
- Open vStorage.
- Red Hat.
- Scality, which has its RING object storage software product.
Exablox, which supported the Toshiba approach when it was announced, is not a member of the KOSP group.
KOSP wants to provide "open-source object storage on next-generation Ethernet-enabled storage devices." In other words, software and systems will see a standard key-value store when accessing Seagate, Toshiba, and Western Digital Ethernet key-value store disks, and they won't need to deal with three separate access protocols and data-moving semantics.
It is blindingly obvious that a common access protocol and semantics are needed for the concept to move forward. There will be APIs, open source libraries, and simulators developed, so that software and system developers can develop new applications and modify existing ones in order to use and manage the key-value store disk drives.
The underlying idea is that a storage array such as a filer adds infrastructure overhead, including a filesystem, between a data-moving application in a server and the target disk drives. It will be more efficient, Seagate and the other diskies say, if server software managed the disk drives directly, using Get and Put object-style semantics. That way the server-app-to-disk-drive stack would be simplified.
It may well be true, but there is a lot of work involved in managing disk drives and tracking which data is stored on which drive. It seems peculiar that object storage startups didn't see the need for this technology and invent it. Instead, three disk drive suppliers – who obviously want to sell disk drives – devised it, with Toshiba and Western Digital's HGST unit following Seagate's lead.
Seagate and HGST have added irons in this fire in that both make disk drive arrays; ClusterStor from Seagate and the Active Archive System from HGST.
There hasn't been any sign of mass adoption of the concept, and only a few JBOD-style enclosures with the drives have been built, for example by Rausch. Seagate says SuperMicro and Dell have agreed to make enclosures for its new drives. The SNIA has a workgroup looking at the concept and developing an Object-Based Storage Device (OSD) specification.
It may well be the case that the key-value store disk drives do find a ready market among hyperscale data center operators needing to store lots of unstructured, variable-type data. They might appreciate the speed and relative simplicity of operating directly-addressed drives in large numbers. But these drives may not find a home in enterprise data centers where files are fine for most people, thank you very much, and buyers appreciate the convenience of having a drive array controller and its software look after their disk drive estate.
The KOSP news has been released without the KOSP organization having produced a website or any collateral material.
Just to avoid confusion, this has nothing to do with the OpenKinect project, which concerns itself with Microsoft Xbox hardware. ®