Physical Storage Hierarchy & Memory Buffers

Welcome to the comprehensive educational guide for CS403 Database Management Systems Lecture 34, focusing on physical storage media, memory hierarchy, and Redundant Arrays of Independent Disks (RAID). In enterprise database management systems, physical storage design fundamentally dictates overall query throughput and transaction latency. Database storage media are classified according to three fundamental engineering characteristics: access speed, cost per unit of data, and storage reliability. The curriculum establishes a core distinction between volatile storage, which loses stored information when power is interrupted, and non-volatile storage, which permanently retains data across power cycles. To balance performance and financial expense, computer systems organize storage into a multi-tiered memory hierarchy. At the apex of the hierarchy are CPU registers and high-speed cache memory constructed from Static Random Access Memory (SRAM). Level 1 (L1) internal caches reside directly on the microprocessor silicon, while larger Level 2 (L2) external caches sit between the CPU and main memory to accelerate repetitive memory requests. Main memory consists of Dynamic Random Access Memory (DRAM), offering access latencies of approximately 50 nanoseconds, though with higher cost per byte than magnetic storage. Non-volatile solid-state storage includes Flash memory (EEPROM), constructed from Floating-Gate Avalanche-Injection Metal Oxide Semiconductor (FAMOS) transistors, providing shock-resistant storage that allows block-level erasure without power consumption. Primary secondary storage relies on magnetic disks, where data is magnetically recorded on spinning platters and accessed via read/write heads. To bridge the severe latency gap between nanosecond DRAM and millisecond magnetic disk seeks, DBMS architectures implement disk caching and buffer pools, storing recently accessed disk blocks in main memory to maximize cache hits. For tertiary backup and archiving, optical disks (CD-ROM, WORM, and Erasable Optical) and sequential-access magnetic tapes provide cost-effective long-term retention. To achieve enterprise-grade fault tolerance and high I/O throughput, systems employ RAID technology. Fundamental to RAID is disk striping, which interleaves data across multiple drives. In I/O-intensive workloads, large stripes allow independent drives to process concurrent queries simultaneously, whereas in data-intensive environments with large records, fine-grained stripes enable parallel read transfers. The course details standard RAID levels: RAID 0 provides striping without redundancy for peak speed but zero fault tolerance; RAID 1 employs disk mirroring for mission-critical redundancy; RAID 2 utilizes Hamming error correction codes; RAID 3 provides byte-level striping with dedicated parity; RAID 4 employs block-level striping with dedicated parity; and RAID 5 implements distributed block-level parity across at least three drives to eliminate the parity write bottleneck. Virtual University students preparing for CS403 finalterm exams must master storage hierarchy trade-offs and RAID architectures using cs403 handouts pdf, solved past papers, and verified MCQs on VUSIGMA.