Disk systems and drives are growing bigger in capacity, and the reliability and performance mechanisms, including RAID, that have worked for smaller disks are beginning to become obsolete. That's the view emerging among some vendors and industry experts.
For one thing, storing a terabyte of data in one place makes the risk of data loss more acute. While the amount of bits on disk is getting larger, the rate of read errors isn't --meaning that each disk has a higher potential for errors than smaller counterparts. Meanwhile, with more capacity on disk and more disks being crammed into arrays, it raises the potential not only for disk failures, but multiple disk failures. Without more advanced RAID algorithms to pick up that slack or a new approach to redundancy, data will be lost.
In response, vendors are starting to take new approaches to providing redundancy and data protection in disk arrays that don't use traditional parity RAID and rebuild approaches.
Products introduced this year, in chronological order include the following:
Atrato Inc.'s Velocity 1000 (V1000): Storage newcomer Atrato began shipping its Velocity system in March, a parallel-access disk array that performs at 11,500 IOPS and holds terabytes of data without requiring drive replacements or hardware maintenance for three to five years. It avoids RAID rebuilds by copying data off a problematic drive, diagnosing whether the drive has truly failed or not, refurbishing bad sectors if possible, and copying the data back to a new drive or the repaired drive. The product also uses small form factor drives, boosting density, and reducing power and cooling requirements.
- No-maintenance disk array aims to eliminate hard drive swapping, March 25
- Storage Soup blog: Still more followup about Atrato
- Wind-powered data center picks Atrato disk arrays, Dec. 10
Xiotech Corp.'s ISE: Based on technology acquired from Seagate [Technology LLC's Advanced Storage Architecture Group] in November 2007, Xiotech's Intelligent Storage Element (ISE) -- a building block for its new Emprise enterprise storage systems -- was unveiled shortly after Atrato's V1000. Xiotech also claims the ISE units eliminate hard drive swapping and guarantee that Emprise arrays will last five years. It avoids RAID rebuilds using a method similar to Atrato's, that of removing corrupted disks from a mix of parallelized drives, and checking for and correcting errors.
However, Xiotech's product is positioned as general-purpose storage, while Atrato is aiming at high-performance computing (HPC) and content-delivery environments. Because of this, Xiotech is selling ISE in much smaller increments than the V1000's ultra-dense configurations. Users can buy an Emprise 5000 unit with as little as 1.5 TB capacity and use it as DAS with a direct upgrade path to the Emprise 7000 SAN system.
- Xiotech flashes self-healing storage systems at SNW, April 8
- Xiotech, Atrato forge market for self-healing storage, April 10
- Xiotech rings up first sale for self-healing Emprise systems, June 16
IBM Corp.'s XIV Storage System: IBM acquired startup XIV Ltd. in January in a deal worth $300 million, and re-released its XIV product in August. The first version of the XIV Storage System released under the IBM brand scales to 180 TB and, unlike most other scale-out storage systems, doesn't use a clustered file system. It also doesn't use RAID, instead mirroring all blocks between two storage nodes within its initial architecture. The company plans to add support for more nodes and higher capacities in future releases.
EMC Corp.'s Atmos: EMC contends that low-cost hardware can provide enough reliability when there are copies of data in enough places. Atmos uses object-based meta data to allow users to set policies that determine where to store information, which services to apply to it, and how many copies should be stored and in which locations. Web services REST and SOAP are built in, as are capabilities such as replication, versioning, compression, data deduplication and disk spin-down. Customers don't have to set up file systems or assign logical unit numbers (LUNs); during setup, they simply answer a few questions to set policy.
ByCast Inc.'s StorageGrid 8: Bycast joined this fray in November with the announcement of StorageGrid 8, an update to its object-based storage software, also previously positioned for data archiving. StorageGrid 8 adds clustered NAS gateways into the system to boost performance, as well as features like virtual server support designed to appeal to cloud service providers. Bycast supports a range of existing storage arrays, including MAID arrays and tape libraries.
Caringo Inc.'s CAStor/Verari Systems Inc.'s DataValet: Caringo's software originally started as content-addressable storage (CAS) for data archiving purposes, but in more recent months has been positioned as an alternative to RAID-based clustered NAS systems. Like EMC Atmos, Caringo software adds data management capabilities such as unique object identifiers, WORM storage, automated lifecycle management according to policy and cryptographic hashes for data integrity.
DataValet consists of blade units that fit into Verari's BladeRack 2. Each unit contains two storage blades that hold up to 12 TB of SATA disk apiece. The nodes are connected via Gigabit Ethernet to each other and to NAS gateway nodes in the BladeRack that connect to clients. Caringo's software is rebranded by Verari to add intelligence to the nodes.
- Users choose Caringo CAS over clustered NAS for nearline storage, May 14
- Blade storage vendor rolls out object-based clustered storage, Nov. 19
Vendors are also adding redundancy to existing RAID systems by grouping commodity server hardware into clusters of redundant nodes. You can read more in our special report on high-performance NAS.