RAID and Mass Storage

RAID Concept

The primary purpose of RAID is to provide high-end storage capabilities and redundant data security for large-scale servers. In a system, RAID is seen as a logical partition, but it is composed of multiple hard disks (at least two). It significantly improves the data throughput of the storage system by simultaneously storing and retrieving data across multiple disks. Many RAID configurations have comprehensive measures for mutual verification/recovery, including direct mirroring backup. This greatly enhances the fault tolerance of RAID systems and improves system stability and redundancy, hence the term “Redundant.”

RAID used to be an exclusive product in the SCSI domain, limited by its technology and cost, which hindered its development in the low-end market. Today, with the increasing maturity of RAID technology and continuous efforts by manufacturers, storage engineers can enjoy relatively more cost-effective IDE-RAID systems. Although IDE-RAID may not match SCSI-RAID in terms of stability and reliability, its performance advantages over single hard drives are quite enticing for many users. In fact, for daily low-intensity operations, IDE-RAID is more than capable.

Similar to modems, RAID can be categorized as fully software-based, semi-software/semi-hardware, or fully hardware-based. Fully software RAID refers to RAID where all functionalities are handled by the operating system (OS) and CPU, without any third-party control/processing (commonly referred to as RAID co-processor) or I/O chip. In this case, all RAID-related tasks are performed by the CPU, resulting in the lowest efficiency among RAID types. Semi-software/semi-hardware RAID primarily lacks its own I/O processing chip, so CPU and driver programs are responsible for these tasks. Additionally, the RAID control/processing chips used in semi-software/semi-hardware RAID generally have limited capabilities and cannot support high RAID levels. Fully hardware RAID encompasses its own RAID control/processing and I/O processing chips, and even includes an array buffer (Array Buffer). It offers the best overall performance and CPU utilization among these three types, but also comes with the highest equipment cost. Early IDE RAID cards and motherboards using HighPoint HPT 368, 370, and PROMISE chips were considered semi-software/semi-hardware RAID, as they lacked dedicated I/O processors. Moreover, the RAID control/processing chips from these two companies had limited capabilities and could not handle complex processing tasks, hence not supporting RAID level 5. A notable example of fully hardware RAID is the AAA-UDMA RAID card produced by Adaptec. It features a dedicated high-level RAID co-processor and Intel 960 specialized I/O processor, fully supporting RAID level 5. It represents the most advanced IDE-RAID product currently available. Table 1 compares typical software RAID and hardware RAID in industry applications.