PHYSICAL DATA ORGANIZATION RAID SYSTEMS Advanced Database Systems

PHYSICAL DATA ORGANIZATION RAID SYSTEMS Advanced Database Systems Dr. Fatemeh Ahmadi-Abkenari September 2013 1

RAID Systems Redundant Array of Independent/ Inexpensive Disks A lower level approach to the performance bottleneck caused by disk I/O Characteristics: v Increasing the availability of data in crash time v Better throughput than a single disk v Concurrent handling of multiple request v Improving the request involving large block of data v Transfer time reduction 2

RAID Systems Terminology: q Mirroring q Striping q Chunks 3 - The unit of data to be spread 2 - Spread of data across an array of disks 1 - One type of redundant storage RAID LEVELS: q BASIC LEVELS (0, 1, 3, 5) q HYBRID LEVELS (01, 10, 03, 30, 50, 100, …) 3

RAID Levels Level 0: § Uses only Striping and no Redundancy § A failure of a single disk is devastating 4

RAID Levels Level 1: § Uses no Striping and only Mirroring § High durability § Expensive for many applications 5

RAID Levels Level 3: § Uses Striping and a type of Redundancy § Chunks are at byte level § The (n+1)st disk is the Parity disk § Parity disk stores the XOR of the corresponding bytes on the other n disks § High transfer rate § Useful for application with real-time requirement 6

RAID Levels Level 3 a b c = a XOR b b c a 0 0 0 0 1 1 1 1 0 1 0 1 a b c d = a XOR b XOR c 0 0 0 1 1 0 1 0 1 0 0 1 1 0 0 1 1

RAID Levels Level 3 A 1: A 2: A 3: 11010110 01111000 11111010 Ap (1 -3): 01010100

RAID Systems Level 5: § Uses Striping and Redundancy § Stores Parity information § No single parity disk 9

Hybrid RAID Systems Level 10: Combination of Level 0 and Level 10 10

Hybrid RAID Systems Level 01 11

Hybrid RAID Systems Level 30 12

Hybrid RAID Systems Level 03 13

Hybrid RAID Systems Level 50 14

Hybrid RAID Systems Level 100 15

Disk Organization Platter Track Sector Cylinder Spindle Arm Assembly Read/Write Head 16

Disk Organization A: Track B: Mathematical Sector C: Disk Sector (Data Unit) D: Cluster

Disk Organization 18

Disk Organization Terminology: q Page: The unit of data transferred with each I/O operation q Block: is a sequence of adjacent sectors on a track q Cache: An array of page-size buffers in main memory q Hit: If an application refer to an item in a page exists in cache Trade-off Choosing the number of sectors in a block (page size) 19

For Further Reading; Database Systems, An application-Oriented Approach Second Edition Chapter 9 Michael Kifer, Arthur Bernstein, Philip M. Lewis Pearson, Addison Wesley Publication 2006