File StructureSNU-OOPSLA Lab1 Appendix A : Designing File Structures for CD-ROM & DVD 서울대학교 컴퓨터공학부 객체지향시스템연구실 SNU-OOPSLA-LAB 교수 김 형

File Structure SNU-OOPSLA Lab 1

Appendix A : Appendix A : Designing File Structures Designing File Structures for CD-ROM & DVD for CD-ROM & DVD

서울대학교 컴퓨터공학부객체지향시스템연구실SNU-OOPSLA-LAB

교수 김 형 주

File Structures by Folk, Zoellick, and Ricarrdi


ObjectivesObjectives

Show how to apply good file structure design principles to develop solutions that are appropriateto this new medium

Describe the directory structure of the CD-ROM file system and show how it grows from the characteristics of the medium


Appendix OutlineAppendix Outline

A.1 Using This Appendix

A.2 Tree Structures on CD-ROM

A.3 Hashed Files on CD-ROM

A.4 The CD-ROM File System

A.5 Summary


Purpose of this AppendixPurpose of this Appendix

Purpose to use the problem of designing file structures for CD-ROM

In this appendix provide a high-level look at how the performance of

CD-ROM affects the design of tree structures hashed indexes directory structures



Tree Structures on CD-ROMTree Structures on CD-ROM

Tree structure are a good way to organize indexes and data on CD-ROM

Avoid seeks is the key strategy in CD-ROM file structure design Sector size is 2k in most CD-ROM Sequential reading performance is moderately fast, for example,

seek of 8k block is relatively efficient than 2k block Block size

Sector size = smallest addressable unit Block size = sector size * constant The large tree structure should usually use at least an 8-Kbytes

block



Tree Structures on CD-ROMTree Structures on CD-ROM: Special Loading Procedures: Special Loading Procedures

Special loading procedures and other consideration B+ tree is commonly used in CD-ROM application

because B+ tree Provide both indexed and sequential access to records Provide very shallow, broad indexes to a set of sequential records Easy to build a two-level index above the sequence set with a

separate loading procedure that builds the tree from bottom up



Tree Structures on CD-ROMTree Structures on CD-ROM: Packing Index: Packing Index

The importance of packing index CD-ROM has relatively large storage,but, sometimes,

we are running out of space because of large documents, images, etc

We need to pack index 100%- full loading procedure

Bottom up organization can pack tree fully



Tree Structures on CD-ROMTree Structures on CD-ROM: Virtual Tree and Secondary Index: Virtual Tree and Secondary Index

Virtual trees and buffering blocks The root of tree should always be buffered in RAM in order

to reduce seek time Buffering below root node should reduce seek time Buffering is most useful when successive accesses to the

tree tend to be clustered in one area Trees as secondary indexes on CD-ROM

CD-ROM applications provide more than one access route to the data on the disc

Secondary index should bind to the target records as tightly as possible



Update of Tree Structures on CD-ROMUpdate of Tree Structures on CD-ROM

Tree structure of CD-ROM will not be updated? One objection is that quite frequently CD-ROM is

reorganized between successive editions of disc Several approaches for this objection

1. Maintain loosely bound records in the source database and transforming them to tightly bound records for publication on CD-ROM

2. Trade off performance on the published disc for decreased costs in producing it



Hashed Files on CD-ROMHashed Files on CD-ROM

Hashing is an excellent way to organize indexes on CD-ROM by its single access retrieval

We should avoid overflow Bucket size

Should be multiple of 2 Kbytes If bucket size < sector size, it would be counterproductive How many sectors into a bucket?

trade-off between seeking and sequential reading Larger buckets require more searching and sequential reading to

find the record in buffer



Packing of Hashed Files on CD-ROMPacking of Hashed Files on CD-ROM

Packing of hashed file should avoid overflow Packing loosely, it causes additional seeking Moderated bucket size

Keep below 60% avoid almost of overflow Packing density of 60% and bucket size of 10

reduce overflow to 1.3 %

reduce the average number of seeks to 1.01



Advantages of Hashed Files on CD-ROMAdvantages of Hashed Files on CD-ROM

Advantages of CD-ROM's read-only status Packing density of index could be up to 100 % We have all keys that are to be hashed at hand We can choose a hash function that provides the

performance we need



CD-ROM File SystemCD-ROM File System

The design goals Support hierarchical directory structures Find and open any one of thousands of files with only

one or two seeks Support the use of generic file names as in “file*.c”

and subdirectory accesses



Two Approaches of CD-ROM File SystemTwo Approaches of CD-ROM File System

Two approaches that were commercially available and tailored to CD-ROM

1. Left-child right sibling tree (Figure A.2) Places the entire directory structure in a single file Works well if the directory structure is small Becomes poor if directory structure is large

2. Hashed index structure(Figure A.3) Creates an index to the file locations by hashing the full

path names of each file Works well if single file is accessed Does very poor job supporting generic file name such as

“file*.c” or directory listing command such as “ls” or “dir”



Hybrid Design of CD-ROM File SystemHybrid Design of CD-ROM File System

Hybrid design - conventional directory structure + hashed index Provide single-seek access to any file Provide the ability to work using generic file names and

commands such as “ls” or “dir” subdirectory retrieval = all file hash processing Approach

Build conventional directory structure

(use a file for each directory) Build index for the subdirectories to solve access problem



Extensions of CD-ROM File SystemExtensions of CD-ROM File System

CD-ROM committee settled on an approach that went one step further They decided to use a special index that take advantage

of that hierarchy of the subdirectories The directory are ordered in the index, parent are

appeared before their children Each child associated with an integer that is a backward

reference to the relative record number(RRN) of the parents

It is good example of a specialized index structure that make use of the hierarchical structure of subdirectories



RRN Parent

0 Root -1

1 Reports 0

2 Letters 0

3 School 1

4 Work 1

5 Personal 2

6 Work 2

Figure A.4 Path Index table of directories


SummarySummary CD-ROM is an electronic publishing medium CD-ROM is built on top of the CD audio The primary disadvantage of CD-ROM is poor seek performance B-tree, B+ tree structures work well on CD-ROM because of their ability

to provide access to many keys with just few seeks Sector size of CD-ROM is 2Kbytes, block size should be multiple of

2Kbytes Larger block is usually advantageous in a tree We had better build index in bottom up fashion because there is no

update, and high index density Hashed index is good choice for CD-ROM because of single seek access CD-ROM file system are directory structure or hashed index Hybrid file system has advantages of two approaches

A.5 Summary


Let’s Review !!!Let’s Review !!!





A.5 Summary

File Structure SNU-OOPSLA-Lab. 20

Introducing “DVD Technology”Introducing “DVD Technology”

서울대학교 컴퓨터 공학부

SNU OOPSLA Lab.

교수 김 형 주

Introducing "DVD Tech."File Structure

SNU-OOPSLA-Lab. 21

Contents• What is DVD?• History from CD to DVD• DVD Capacity• CD v.s. DVD• File Structure

• Track Structure• Sector Structure• Block Structure• Track Buffer Structure• Tree Structure• Hashed Files

• DVD Video


SNU-OOPSLA-Lab. 22

What is DVD?What is DVD? DVD

Digital Video disk (DVD-Video) Digital Versatile disk (DVD-ROM)

In september 1995 As a movie-playback format As a computer-ROM format

Next-Generation optical disc storage tech. will replace audio-CD,videotape,laserdisk, CD-ROM,etc.


SNU-OOPSLA-Lab. 23

The history from CD to DVDThe history from CD to DVD

1980, Sony & Philips --> CD-Audio 1985, Sony & Philips --> CD-ROM 1989, Sony & Philips --> CD-I 1990, Sony & Philips --> CD-R 1995, --> CD-E 1995, september --> DVD


SNU-OOPSLA-Lab. 24

DVD CapacityDVD Capacity

Single-sided DVD5 ( 4.7 GB/single-layer ) DVD9 ( 8.5 GB/dual-layer )

Double-sided DVD10 ( 9.4 = 4.7x2 GB/dual-layer ) DVD18 ( 17 = 8.5x2 GB/dual-layer )

Write-Once DVD-R ( 3.8 GB/side )

Overwrite DVD-RAM ( more than 2.6 GB/side )


SNU-OOPSLA-Lab. 25

0.6mm

0.6mm

0.6mm

0.6mm

reflexive-layer substrate

semi-transmissive-layer reflexive-layer

(a)

(b)

(a) Single sided, single layer (b) Single sided, dual layer

(gold-layer) (silver-layer)


SNU-OOPSLA-Lab. 26

CD vs. DVDCD vs. DVD

Laser-Beam CD --> infrared light ( 780nm ) DVD --> red light ( 635-650nm )

Capacity CD --> maximum 680MB DVD --> maximum 17GB( 25 times of CD )

Reference Speed CD --> 1.2m/sec. CLV DVD --> 4.0m/sec. CLV


SNU-OOPSLA-Lab. 27

Track StructureTrack Structure

Legend I Lead-in area (leader space near edge of disc)

D Data area (contains actual data)

O Lead-out area(leader space near edge of disc)

X Unusable area (edge or donut hole)

M Middle area (interlayer lead-in/out)

B Dummy-bonded layer

(to make disc 1.2mm thick instead of 0.6mm)


SNU-OOPSLA-Lab. 28

BBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBXX I I I DDDDDDDDDDDDDDDDDDDDDOOOXX

Single layer disc :

direction: continuous spiral from inside to outside of disc.

reference axis outer edge of disc


SNU-OOPSLA-Lab. 29

Dual layer disc :

(A) Parallel track path (for computer CD-ROM use) Direction : same for both layers.

(B) Opposite track path (for movies) Direction : opposite directions (Since the reference beam and angular velocities are the same at the layer transition point, the delay comes from refocusing.

This permits seamless transition for movie playback.)


SNU-OOPSLA-Lab. 30

XX I I I DDDDDDDDDDDDDDDDDDDDDOOOXX layer 1 XX I I I DDDDDDDDDDDDDDDDDDDDDOOOXX layer 0

XX I I I DDDDDDDDDDDDDDDDDDDDDOOOXX layer 1 XX I I I DDDDDDDDDDDDDDDDDDDDDOOOXX layer 0

reference axis outer edge of disc

(A)

(B)

(A) parallel track-path (B) Opposite track-path


SNU-OOPSLA-Lab. 31

Sector StructureSector Structure

2064 bytes/sector organized into 12 rows, each with 172bytes first row starts with 12B sector header (ID,IEC,Reserved bytes) final row is punctuated with 4B (EDC bytes)

172 x 12 = 2064 bytes/sector 12 rows

172bytes/rows


SNU-OOPSLA-Lab. 32

Row Fields within row

0 ID(4B) IEC(2B) RESERVED(6B) Main data(160B : D[0]-D[159] )1 Main data( 172B : D[ 160]-D[ 331] ) 2 Main data( 172B : D[ 332]-D[ 503] )3 Main data( 172B : D[ 504]-D[ 675] )4 Main data( 172B : D[ 676]-D[ 847] )5 Main data( 172B : D[ 848]-D[1019] )6 Main data( 172B : D[1020]-D[1191] )7 Main data( 172B : D[1192]-D[1363] )8 Main data( 172B : D[1364]-D[1535] )9 Main data( 172B : D[1536]-D[1707] )10 Main data( 172B : D[1708]-D[1879] )11 Main data( 168B : D[1880]-D[2047] ) EDC(4B)

ID : Identification Data ( 32bit sector number)IEC : ID Error CorrectionEDC : Error Detection Code


SNU-OOPSLA-Lab. 33

Block StructureBlock Structure

To combat burst error, 16 sectors are interleaved together ( 16 sectors * 12 rows/sector = 192 rows )

Error correction byes are concatenated 10bytes at the end of each row 16 rows at the end of the block


SNU-OOPSLA-Lab. 34

172bytes 10B

192 rows

16rows

Data Block

Error correction byes

payload/block = 172 x 192

182 x 208x 100 = 87 %


SNU-OOPSLA-Lab. 35

Track BufferTrack Buffer

The size of the track buffer is left to implementation, although the minimum recommended size is 2 MB

( Track buffer > Tmax * VBRmax = 0.104 sec * 10.08 MB/sec = 1.04832MB )

Tmax : max latency of one disc evolution VBRmax : max mux rate for any program


SNU-OOPSLA-Lab. 36

( Input stream to Track Buffer )

[n-2][n-1][n] .......... track jump ....... [m][m+1][m+2]

T

( no data transfer during discontinuity )

( Output stream from Track Buffer )

[n-2][n-1][n][m][m+1][m+2]

( no apparent discontinuity )

( Initial buffer delay introduced by track buffer )


SNU-OOPSLA-Lab. 37

Tree StructureTree Structure

Tree Structure are a good way to organize indexes & data on DVD

Design Issue Block size of B-tree & B+ tree Memory size of buffering blocks Loading procedure of B+ tree implementation Access mechanism of primary index & secondary index


SNU-OOPSLA-Lab. 38

Tree Structure(2)Tree Structure(2)

Block Size if block size is big, then can provide access to a large number of record in only

a few seeks. >= sector size DVD-ROM’s sequential reading performance is moderately fast than seeking

performance, so it is benefit to use a block composed of several sectors.


SNU-OOPSLA-Lab. 39

Tree Structure(3)Tree Structure(3) B+ tree

provides both indexed & sequential access to records provides very shallow, broad indexed to a set of sequenced records. provides access to millions of records with an index that is only two levels deep. if the root of index is kept in RAM, then the cost of searching is reduced to a

single seek. 100% - full loading procedure can be designed using bottom-up


SNU-OOPSLA-Lab. 40

Hashed FilesHashed Files

Hashing is an excellent way to organize indexes on DVD-ROM with a single access retrieval.

Design Issue Bucket Size Packing density for the hashed index Hash Function


SNU-OOPSLA-Lab. 41

Hashed Files(2)Hashed Files(2)

Bucket Size >= sector size multiple of sector size trade-off between seeking & sequential reading keeping the packing density below 60% will tend to avoid overflow almost all

the time. Hash Function

Hash function-fitting effort is worthwhile because of the asymmetric nature of writing & reading DVD-ROM


SNU-OOPSLA-Lab. 42

DVD Video FeaturesDVD Video Features

Over 2 hours of high-quality digital video (over 8 on a DS,DL disc) Support wide screen movies & standard or widescreen TVs ( 4:3 & 16:9 aspect ratio

s ) Up to 8 tracks of digital audio Up to 32 subtitle/karaoke tracks Up to 9 camera angles Multilingual identifying text for title name, album name, song name, actors, etc.


SNU-OOPSLA-Lab. 43

DVD Video Encoding DataDVD Video Encoding Data

Encoding Image MPEG-2 compression ( developed by the Motion Pictures Experts Group ) High-Resolution ( better than CD,LD 3-times better than Video tape )

Encoding Sound Dolby Digital surround AC-3 sound compression ( support five sound channel plus subwoofer channel => left, center, right, rear-left, rear-right channel )


SNU-OOPSLA-Lab. 44

DVD Video CardDVD Video Card

Encoding Data

MPEG-2( Image )

AC-3 ( sound )

Decoder

Decoding Data


SNU-OOPSLA-Lab. 45

Let’s Review !!!• What is DVD?• History from CD to DVD• DVD Capacity• CD v.s. DVD• File Structure

• Track Structure• Sector Structure• Block Structure• Track Buffer Structure• Tree Structure• Hashed Files

• DVD Video

Documents

File StructureSNU-OOPSLA Lab1 Appendix A : Designing File Structures for CD-ROM & DVD 서울대학교 컴퓨터공학부 객체지향시스템연구실 SNU-OOPSLA-LAB 교수 김 형