Most corporations use relatively isolated and expensive disk subsystems for primary storage, and they protect this data with tape back-up systems that are stored offsite for disaster-recovery purposes.
A new storage system architecture called Redundant Array of Inexpensive Nodes (RAIN) surpasses this traditional storage architecture by offering data-storage and protection systems that are more distributed, shareable and scalable. RAIN systems also are less expensive than traditional systems.
RAIN is an open architecture approach that combines standard, off-the-shelf computing and networking hardware with highly intelligent management software. This combination lets a host of storage and data-protection applications be cost-effectively deployed across a grid of devices that are highly available and self-healing.
RAIN-based storage and protection systems consist of:
- RAIN nodes: These hardware components are 1U servers that provide about 1 terabyte of serial ATA (SATA) disk storage capacity, standard Ethernet networking and CPU processing power to run RAIN and data management software. Data is stored and protected reliably among multiple RAIN nodes instead of within a single storage subsystem with its own redundant power, cooling and hot-swap disk-drive hardware.
- IP-based internetworking: RAIN nodes are physically interconnected using standard IP-based LANs, metropolitan-area networks (MAN) and/or WANs. This lets administrators create an integrated storage and protection grid of RAIN nodes across multiple data centers. With MAN and WAN connectivity, RAIN nodes can protect local data while offering off-site protection for data created at other data centers.
- RAIN management software: This software lets RAIN nodes continuously communicate their assets, capacity, performance and health among themselves. RAIN management software automatically can detect the presence of new RAIN nodes on a new network, and these nodes are self-configuring.
The management software creates virtual pools of storage and protection capacity without administrative intervention. It also manages all recovery operations related to one or more RAIN nodes becoming unavailable because of RAIN node or network failures. RAIN nodes do not require immediate replacement upon component failure because lost data is automatically replicated among the surviving RAIN nodes in the grid.
- Information life-cycle management software: This software replaces traditional snapshot, back-up and mirroring data-management tools with innovative virtualization, compression, versioning, encryption, self-healing integrity checking and correcting, retention and replication algorithms. Information life-cycle management software increases the overall reliability of lower-cost SATA disk drives by replicating data among multiple RAIN nodes.
A grid of RAIN nodes also can adapt to changing application workloads by load-balancing data across nodes based on utilization or storage capacity.
In a RAIN-based storage system, each RAIN node regularly checks all its own files. The combination of hundreds of RAIN nodes forms a powerful parallel data-management grid - one that is much more powerful than today's independent protection architectures. When file corruption is detected, the associated RAIN node initiates a replication request to all other RAIN nodes, which verify their own replicas and work collectively to replace the defective file.
Grids of RAIN nodes will replace existing isolated data-storage systems. Low-cost, high-performance disk drives, CPUs and IP networking make this evolution possible. In addition, businesses are demanding simplified, lower-cost, site disaster-recovery systems and faster and more reliable back-up and restore processes.
By executing information life-cycle management applications across hundreds of powerful, internetworked storage and compute RAIN nodes, RAIN systems will deliver unprecedented long-term data availability, cost-effective and rapid site disaster recovery, and automated onsite and offsite data back-up protection.