1. Chöông 10: Heä Thoáng File
10.B
Hieän thöïc heä thoáng file vaø thö muïc
Caùc phöông phaùp quaûn lyù khoâng gian troáng

2. Sô ñoà boá trí (layout) heä thoáng file (1/4)
Toå chöùc khoâng gian ñóa (maùy tính caù nhaân – PC)
MBR
From
Several major kinds of boot sectors could be encountered on IBM PC compatible hard disks, floppy disks and similar storage devices:
A master boot record (MBR) is the first sector of a data storage device that has been partitioned. The MBR sector may contain code to locate the active partition and invoke its Volume Boot Record.
A volume boot record is the first sector of a data storage device that has not been partitioned, or the first sector of an individual partition on a data storage device that has been partitioned. It may contain code to load and invoke an operating system (or other standalone program) installed on that device or within that partition.
Partition control block
FCB’s
i-nodes

3. Sô ñoà boá trí heä thoáng file (2/4)
Partition control block
löu soá löôïng block trong partition, kích thöôùc block, soá löôïng free block hieän thôøi vaø caùc con troû chæ ñeán chuùng,…
löu soá löôïng free FCB hieän thôøi vaø caùc con troû chæ ñeán chuùng,…
Ví duï “superblock” trong UNIX File System
File control block (FCB): moãi file ñöôïc quaûn lyù thoâng qua FCB cuûa noù
löu caùc thoâng tin veà file, keå caû caùc con troû chæ ñeán caùc data block cuûa noù
Ví duï “i-node” trong UNIX File System:

4. Sô ñoà boá trí heä thoáng file (3/4)
Layout cuûa moät partition chöùa heä thoáng file UNIX

5. Sô ñoà boá trí heä thoáng file (4/4)
FAT duøng ñeå chæ baûng FAT vaø cuõng duøng ñeå chæ heä thoáng file
Layout cuûa moät partition chöùa heä thoáng file FAT
Boot sector
FAT
Root directory
Data blocks

6. VFS (Virtual File System)
Cung caáp giao dieän ñoàng nhaát cho öùng duïng ñoäc laäp vôùi file system cuï theå  nhieàu file system khaùc nhau trong cuøng heä thoáng
open, read, write, opendir,…
disk
partition
ext2 file system
FAT file system
NFS file system
ÖÙng duïng
VFS
: goïi haøm/thuû tuïc
switch

7. VFS (Virtual File System)
Vò trí cuûa VFS vaø file system software trong I/O call path (Linux)
ÖÙng duïng trong user space
VFS
ext2_file_write()
File system cuï theå, ôû ñaây ext2
Ñònh thôøi cho thieát bò block
Device driver
from Lunde

8. Hieän thöïc file
Caáp phaùt khoâng gian löu tröõ cho file/directory, muïc tieâu:
söû duïng khoâng gian ñóa höõu hieäu
truy caäp file nhanh
Neáu soá löôïng vaø kích thöôùc file khoâng thay ñoåi ñoäng thì hieän thöïc file nhö theá naøo?
Caùc phöông phaùp caáp phaùt phoå bieán
Caáp phaùt lieân tuïc (contiguous allocation)
Caáp phaùt theo danh saùch lieân keát (linked list allocation)
Caáp phaùt duøng chæ muïc (indexed allocation)

9. Caáp phaùt lieân tuïc Thôøi gian di chuyeån ñaàu ñoïc?
Coù theå truy xuaát ngaãu nhieân moät block cuûa file: block nr = start + block offset
Phaân maûnh ngoaïi
Coù theå gaëp khoù khaên khi taïo file môùi vaø khi caàn theâm block cho file
ÖÙng duïng: ISO-9660 (CDROM)

10. Caáp phaùt theo danh saùch lieân keát (1/2)
pointer -1
data
layout cuûa block -1

11. Caáp phaùt theo danh saùch lieân keát (2/2)
Öu ñieåm
Deã daøng theâm block cho file khi caàn
Quaûn lyù khoâng gian troáng baèng danh saùch lieân keát
Khoâng coù phaân maûnh ngoaïi
Nhöôïc ñieåm
Chæ truy xuaát hieäu quaû ñoái vôùi sequential-access file
Toán khoâng gian löu tröõ caùc con troû
Ñoä tin caäy: pointer trong block coù theå bò hoûng

12. FAT – moät hieän thöïc cuûa caáp phaùt theo danh saùch lieân keát:
Nhöng khoâng löu con troû ñeán file block tieáp theo trong block chöùa döõ lieäu file
FAT (File Allocation Table)
Moãi block ñóa ñöôïc töôïng tröng bôûi moät entry trong FAT
FAT entry vôùi index i töôïng tröng block coù block nr i
FAT entry chöùa block nr keá tieáp trong file, neáu file goàm nhieàu block
no. of disk blocks - 1

13. Caáp phaùt duøng chæ muïc (1/2)
Baûng index (index block)
chöùa ñòa chæ caùc block cuûa file
thöù töï caùc ñòa chæ cuõng laø thöù töï caùc block cuûa file

14. Caáp phaùt duøng chæ muïc (2/2)
Öu ñieåm
Random vaø sequential access
Khoâng xaûy ra phaân maûnh ngoaïi
Khuyeát ñieåm
Toán khoâng gian löu tröõ baûng index duø file coù kích thöôùc chæ vaøi block
Vaán ñeà: Kích thöôùc moät file coù theå nhoû nhöng cuõng coù theå raát lôùn. Kích thöôùc index block bao nhieâu laø phuø hôïp?
Giaûi quyeát: multilevel index  i-node

15. i-node – moät hieän thöïc cuûa index block
UNIX v7 i-node: 13 pointer
Linux ext2 i-node: 15 pointer

16. Hieän thöïc file duøng i-node
File (user view)
Ví duï

17. Hieän thöïc thö muïc
Thö muïc ñöôïc duøng ñeå chöùa baûng aùnh xaï töø teân file (chuoãi kyù töï ASCII) ñeán thoâng tin caàn thieát ñeå ñònh vò caùc block döõ lieäu cuûa file
Toå chöùc thö muïc
Danh saùch tuyeán tính (array hay linear list), baûng baêm,…
FAT file system
UNIX file system
first block nr
i-node

18. Duyeät path name ñeå laáy block nr cuûa file
Ví duï: Xaùc ñònh caùc block döõ lieäu cuûa file /a/b/c
i-node of /
(the 1st one of i-list, by convention)

19. i-node: chia seû file (1/2)
Thö muïc
i-node 27
i-node 51

20. i-node: chia seû file (2/2)
Thö muïc
i-node 27
i-node 51
i-node 27

21. Quaûn lyù khoâng gian troáng
Caùc phöông phaùp
Bit vector (bit map)
Linked list
Grouping
Counting

22. Phöông phaùp bit vector (bit map)
Ví duï:
bit vector … 
block 0, 1 troáng
block 2, 3, 4, 5 ñaõ ñöôïc caáp
block 6, 7 troáng … 1 2
n - 1
bit[ i ] =
0  block i coøn troáng
1  block i ñaõ ñöôïc caáp
Öu ñieåm: Ñôn giaûn vaø hieäu quaû khi caàn tìm khoái troáng ñaàu tieân hoaëc chuoãi khoái troáng lieân tuïc
Thao taùc treân bit
Khuyeát ñieåm: Caàn khoâng gian löu tröõ. Ví duï
Kích thöôùc block = 212 byte
Kích thöôùc ñóa = 230 byte
n = 230/212 = 218 bit (32 KB)

23. Phöông phaùp duøng linked list
Lieân keát caùc khoái troáng vôùi nhau
Chæ caàn giöõ con troû ñeán khoái nhôù troáng ñaàu tieân treân ñóa, coù theå cache trong boä nhôù chính ñeå taêng toác
Öu ñieåm: Ít laõng phí khoâng gian ñóa
Khuyeát ñieåm: Khoâng hieäu quaû; trong tröôøng hôïp xaáu nhaát phaûi duyeät toaøn boä ñóa ñeå tìm khoâng gian troáng lieân tuïc (tröôøng hôïp caáp phaùt lieân tuïc)

24. Grouping vaø counting (1/2)
Phöông phaùp grouping
Caùc ñòa chæ cuûa n khoái troáng ñöôïc löu trong moät khoái troáng ban ñaàu.
Khoái nhôù thöù n chöùa caùc ñòa chæ cuûa n khoái nhôù troáng keá tieáp.
Nhaän xeùt: grouping laø môû roäng cuûa phöông phaùp duøng linked list
(laáy n = 1)
Phöông phaùp counting
Toå chöùc baûng chæ muïc
moãi entry: ñòa chæ cuûa khoái troáng ñaàu tieân trong nhoùm khoái troáng lieân tuïc vaø moät soá ñeám soá löôïng khoái troáng.
Coù theå caáp phaùt hoaëc thu hoài ñoàng thôøi nhieàu khoái nhôù lieân tuïc.

25. Grouping vaø counting (2/2)
Ví duï: Phương phaùp linked list
Phương phaùp grouping: n = Block 2 löu 3, 4, Block 5 löu 8, 9, Block 10 löu 11, 12, Block 13 löu 17, 28, Block 25 löu 26, 27
Phöông phaùp counting: noäi dung index block

26. Journaling file system
Ghi nhaän caùc laàn caäp nhaät treân file system thaønh caùc giao taùc (transaction)
Moïi transaction ñeàu phaûi ñöôïc ghi nhaän trong log file
Moät transaction ñöôïc xem laø hoaøn taát (commit)  ñaõ ñöôïc ghi nhaän ñaày ñuû trong log file (luùc naøy, file system coù theå chöa ñöôïc caäp nhaät)
Khi file system ñöôïc caäp nhaät vôùi ñaày ñuû moïi taùc vuï trong transaction thì transaction seõ ñöôïc xoùa ñi trong log file
Neáu file system bò hoûng  heä ñieàu haønh döïa vaøo caùc transaction trong log file ñeå söûa chöõa
Tham khaûo theâm Linux-ext3, JFS, NTFS
The concept of journaling in file systems, or Write Ahead Logging (WAL) as it is called in database systems, was introduced to improve speed and reliability during crash recovery. The journal feature of ext4 (introduced in ext3, see [23]) is a circular buffer in which disk writes are logged before they are written to the main file system. This allows recovery because at all times the set of blocks which might be inconsistent is known, and the blocks are stored in a consistent state; either the new version in the journal, or the old version on the main file system.
The journaling feature has a large impact on the patterns of disk writes as seen from the I/O scheduler. While full data journaling increases the amount of writes by a factor of two, because data is written to both the journal and the main file system, the write patterns become much nicer.
The writes to the data journal, which may be either a pre-allocated file (represented using an inode number stored in the superblock–normally and extents based file with inode number 8), or a separate block device, are very sequential due to the nature of the log (append-only circular buffer).
Unless the file system is very small or a custom block or journal size is used, the journal will be 128 MiB large and stored in the middle of the file system, in a non-sparse block group without any inode tables and bitmaps.
After data is written to the journal, the file system may delay writing the data to the main file system longer than it could have done without journaling, especially when system calls such as fsync are used.
As shown in [24], the impact of journaling depends on the workload. For sequential writes, full data journaling may halve the bandwidth, while metadata journaling has low cost. On the other hand, with highly random workloads data journaling may give a factor of 10 in performance gain, while plain metadata journaling again has little to no overhead.
[23] Stephen C. Tweedie. Journaling the linux ext2fs filesystem. In The Fourth Annual Linux Expo, Durham, North Carolina, May Available from: [cited May 2009].
[24] Vijayan Prabhakaran, Andrea C. Arpaci-Dusseau, and Remzi H. Arpaci-Dusseau. Analysis and evolution of journaling file systems. In ATEC ’05: Proceedings of the annual conference on USENIX Annual Technical Conference, pages 8–8, Berkeley, CA, USA, USENIX Association.

27. Phuï luïc

28. MS-DOS File System
MS-DOS directory entry

29. Windows 98 File System (1/3)
Môû roäng MS-DOS directory entry ñeå duøng trong Windows 98
Bytes

30. UNIX V7 File System (1/2)
Moät UNIX V7 directory entry
i-node
number

31. UNIX V7 File System (2/2)
Caùc böôùc ñeå doø tìm /usr/ast/mbox