Disk array controller

Lore & Background

Disk array controllers have front-end and back-end interfaces. The back-end communicates with the controlled disks using protocols such as ATA, SATA, SCSI, FC, or SAS. The front-end communicates with a computer's host adapter, using protocols like ATA, SATA, SCSI, FC, or less common dedicated protocols such as FICON/ESCON, iSCSI, HyperSCSI, ATA over Ethernet, or InfiniBand. A single controller may use different protocols for front-end and back-end; many enterprise controllers use FC on the front-end and SATA on the back-end.

In enterprise architectures, disk array controllers—sometimes called storage processors (SPs)—are parts of physically independent enclosures, such as disk arrays in a storage area network (SAN) or network-attached storage (NAS) servers. These external disk arrays are typically purchased as integrated subsystems of RAID controllers, disk drives, power supplies, and management software. The controllers provide advanced functionality including automatic failover, long-running operations without downtime, forming new RAID sets, reconstructing degraded RAID sets, adding or removing disks online, partitioning RAID sets into volumes/LUNs, snapshots, business continuance volumes (BCV), and replication with a remote controller.

Simple disk array controllers may fit inside a computer as a PCI/PCIe expansion card or be built onto a motherboard, often providing host bus adapter (HBA) functionality themselves and sometimes called a RAID adapter. Intel integrated their Matrix RAID controller in upmarket motherboards, controlling 4 devices plus 2 additional SATA connectors for 6 total SATA connections at 3 Gbit/s each, with one IDE connector for backward compatibility.

Reader's Guide

The disk array controller's significance lies in its role as the bridge between physical storage and logical presentation, enabling RAID functionality that provides fault tolerance, performance, and capacity management. Historically, hardware RAID controllers initially required expensive Parallel SCSI hard drives and targeted the server and high-end computing market, with SCSI advantages including support for up to 15 devices per bus, independent data transfers, hot-swapping, and higher MTBF. Around the introduction of ATAPI-4 and Ultra-DMA-Mode, the first ATA RAID controllers appeared as PCI expansion cards, bringing RAID to the consumer market for users seeking fault tolerance without investing in expensive SCSI drives. However, most ATA RAID controllers lacked dedicated buffers or high-performance XOR hardware, resulting in relatively poor performance compared to SCSI RAID controllers, and data safety suffered without battery backup for writes interrupted by power outages.

Operating system support varies widely. Because hardware RAID controllers present assembled RAID volumes, operating systems are not strictly required to implement complete configuration for each controller. Basic features are often implemented in open-source drivers, with extended features provided through binary blobs from manufacturers. RAID controllers are typically configured through card BIOS before booting, and proprietary configuration utilities are available from each manufacturer. Unlike Ethernet network interface controllers, which can be configured through common OS paradigms without third-party tools, each RAID controller manufacturer usually provides proprietary software for each supported OS, creating vendor lock-in and potential reliability issues. Some operating systems have implemented generic frameworks for interfacing with any RAID controller, such as OpenBSD's bio(4) pseudo-device driver and bioctl utility from 2005, which provide volume status, LED/alarm/hotspare control, and health monitoring sensors, keeping the feature set minimal so each controller can be supported uniformly.

Did You Know?

• A disk array controller is often ambiguously shortened to 'disk controller,' which can also refer to internal drive circuitry.
• Many enterprise controllers use Fibre Channel on the front-end and SATA on the back-end.
• Intel integrated a Matrix RAID controller in upmarket motherboards providing 6 SATA connections at 3 Gbit/s each.
• OpenBSD introduced the bio(4) pseudo-device driver and bioctl utility in 2005 for generic RAID controller monitoring.