Dell OEMs Data Domain and Celerra
Adding its own object storage technology
Dell is broadening its storage product range by OEMing EMC's Celerra and Data Domain products, and developing its own object storage product.
The background to this is the continuing and dramatic rise in the amount of semi- and unstructured information. This leads to a sheer storage capacity problem and to infrastructure problems, particularly when organisations need to respond to sudden and unpredictable changes in IT demand.
Dell is responding to that by increasing the ability for its customers to virtualise their IT environments, manage them more efficiently with infrastructure products, and equip their data centres with both servers and storage better suited to what it calls the virtual era.
The Reg covers the server, cloud and infrastructure parts of today's Dell announcement set elsewhere; here we concentrate on the storage which focuses on the efficient storage of billions of files and objects.
Dell is introducing three new storage product sets: the DX object store; DD deduplication systems; and the NS unified file and block storage systems Both the DD and NS products are OEM'd from EMC whereas the DX is not. It appears that Dell decided not to take EMC's Centera object storage product set.
DX Object Storage Platform
Paul Prince, a director in Dell's chief technology officer's organisation, said this was like iSCSI. There Dell saw a strongly growing market and decided it needed its own IP. Consequently it went out and bought EqualLogic.
Dell thinks that the growth in unstructured data will mean that potentially billions of files need to be stored in a searchable address space and that object storage is the best way to do that. It says DX customers will be able to access, store and distribute the billions of files or other digital content, from archiving all the way to the cloud.
The DX Object Storage Platform is a coming product and details are somewhat scarce. We know that it is based on one or more Cluster Services Node (CSN) management front end connected by 1GBitE to DX6012s storage nodes which can be added and added to build multi-petabyte storage capabilities with no architectural limit on the number of nodes.
The base DX technology components come from unidentified partners; Caringo thought to be one, with Dell taking the best of breed parts and integrating them to produce its own object storage technology. Data is stored with metadata which is used to automatically manage the length and location of content storage. The company did not say if it was used hash addressing to do this.
The CSN boxes store an index or map of all the objects in the storage nodes and they store the objects in a single flat and potentially enormous address space. They are 2U enclosures with a single X86 CPU, six hard drives offering up to 6TB max capacity, 12GB of RAM, and 4 1GbitE ports. They have redundant power supplies and an iDRAC Express management card.
These CSNs have the centralised management interface, carry out network services and contain a content router
The DX6012s SNs also come in 2U enclosures, this time with up to 12 hard drives offering 3,6, 12 or 24TB of capacity. There is a single X86 CPU and 3, 6, 12 or 24GB of RAM plus two 2GbitE ports and redundant power supplies.
The nodes are self-managing and carry out load-balancing, storage optimisation and power conservation. They also carry out background data integrity health checks on their stored content, both original data and replicas.
The DX 6000 has policy-based replication to geographically separate sites, with policies driven by administrator-defined object metadata. Multiple replication and distribution topologies are supported such as 1:1, 1:Many, Many:1, Many:Many.
Dell says the DX system will feature automated, policy-based retention and deletion and be affordable. It will offer various data types: immutable, mutable, and conditional mutable.
The DX is claimed to be highly expandable through a peer-scaling design. There will be multiple options for scaling capacity. It will also, Dell claims, be possible to seamlessly add future storage technologies.
This is likely to be made possible by the storage nodes having standard interfaces to the management nodes. Dell did confirm that SATA drives will be used in the storage nodes but said that other drive types or storage tiers, such as Write-Once-Read-Many (WORM) would be available as well.
The interconnect between the storage and management nodes was not identified and nor was the general speeds and feeds type information. Dell says that setup will use wizards and will not require LUNs or RAID groups. The hardware is industry-standard, and it's the DX software where the value and IP lies.
It's likely that some kind of data moving function across storage tiers will be provided.
Dell is working to build an eco-system of horizontal and vertical software suppliers to provide data ingest and access capabilities that use the DX as their object storage platform. The base protocol is HTTP and Dell is making its DX APIs available in a Software Development Kit (SDK) to partners, such as Acuo, Bridgehead Software, CommVault, EMC, Iron Mountain, Symantec and others.