Computer Research Laboratory, EECS, KAIST 1

EE311 OS

April 6, 2010Prof. Kyu Ho Park

http://core.kaist.ac.kr

Lecture 10:File Systems Implementation

Computer Research Laboratory, EECS, KAIST 2

File System Implementation

New Trend- Storage Class Memory(SCM)

File-System Structure File-System Implementation Directory Implementation Allocation Methods Free-Space Management

Computer Research Laboratory, EECS, KAIST

New Memory Architecture

Storage Class Memory(SCM)

3

Computer Research Laboratory, EECS, KAIST

System EvolutionCPU-RAM-Disk

CPU-RAM-SSD-Disk

CPU-RAM-SCM-Disk

4

Computer Research Laboratory, EECS, KAIST

Hierarchy of Latency Freitas &Wilcke, IBM J.Res&Dev,2008

Disk 10E7-10E8 CPU cyclesSCM 10E3DRAM 10E2L2,L3 cache 10-100L1 1

5

Computer Research Laboratory, EECS, KAIST

2009 2011 2013 2015 2017

NAND Flash-SLC* 0.0081

0.0052

0.0031

0.0021

0.0013

NAND Flash-MLC* 0.0041

0.0013

0.0008

0.0005

0.0003

PCRAM(nMOSFET)-SLC* 0.0254

0.0123

0.0069

0.0036

0.0024

PCRAM(nMOSFET)-MLC* 0.0127

0.0061

0.0017

0.0009

0.0006

DRAM Cell density 0.0153

0.0096

0.0061

0.0038

0.0024

Flash write/erase cycles 1E+05

1E+05

1E+05

1E+05

1E+05

PCRAM write/erase cycles 1E+10

1E+10

1E+12

1E+15

1E+15

Flash SLC/MLC data retention (years) 10-20 10-20 10-20 10-20 10-20

PCRAM SLC/MLC data retention (years)

>10 >10 >10 >10 >10

)/( 2 bitm

)/( 2 bitm

)/( 2 bitm

)/( 2 bitm

)/( 2 bitm

*SLC-Single level Cell, MLC-Multi Level Cell

Roadmap for Memory Technology

D.Roberts, T,Kgil, and T.Mudge, EDAA 2009

Computer Research Laboratory, EECS, KAIST

New Memory ArchitecturesNVRAMOS 2009

An Empirical Study using NVRAM

Performance/Energy tradeoffs on NVRAM

Operating System Support for NVRAM Green data center with NVRAM

CPU

SCM

CPU

SCM

NOR, NAND Flash

CPU

RAM

SCM

CPU

RAM

NOR, NAND Flash

RAM-Flash RAM-SCM SCM - Flash

SCM - Only

< Best power efficiency >

< Best Performance >

Computer Research Laboratory, EECS, KAIST

New Memory ArchitecturesNVRAMOS 2009

I/O bound job SCM is best

Memory bound job DRAM is best

CPU bound job Little impacts

Computer Research Laboratory, EECS, KAIST

SCM only SystemNVRAMOS 2009

SCM Great potential to reduce energy

consumption SCM as memory cause performance

degradationCPU

SCM

SCM - Only

Computer Research Laboratory, EECS, KAIST

Operating System support for SCM

NVRAMOS 2009

Operating System support for SCM Unified file object and memory objects Eliminate redundant I/O accesses mmap like operations for all

memory/disk data

Buffer cache

File Disk- Twice accesses for same data to update

SCMFilememor

y

Computer Research Laboratory, EECS, KAIST

Adaptive Context SwitchNVRAMOS 2009

Context Switch on Block devices Fast block devices with SCM

No need to context switch, which takes 100us

Keep use whole schedule quantum 5~10% performance improvements Shared object accesses in multiple

threads cause performance degradation/malfunctions(Why?)

CPU

RAM

SCM

I/O access less than 1ms

Computer Research Laboratory, EECS, KAIST 12

Objectives

To describe the details of implementing local file systems and directory structuressupplementary to Lecture 9.

Computer Research Laboratory, EECS, KAIST 13

File-System Structure File structure

Logical storage unit Collection of related information

File system resides on secondary storage (disks)

File system organized into layers File control block – storage structure

consisting of information about a file

Computer Research Laboratory, EECS, KAIST

File Systems How the file system can be looked to

the user; Define a file and its attributes , the

operations allowed on a file, and the directory structure for organizing a file.

Algorithms and data structures to map the logical file system to the physical disk.

14

Computer Research Laboratory, EECS, KAIST 15

Layered File System

Computer Research Laboratory, EECS, KAIST

Layered File System I/O control;

It consists of device drivers and interrupt handlers to transfer information between the main memory and the disk system.Device drivers: Its input consists of high-level commands.

“get block 1000” “H/W specific

instructions”

16

Device Driver

Computer Research Laboratory, EECS, KAIST

Basic File System Layer It issues generic commands to the

appropriate device drivers to read and write physical blocks on the disk.

Each physical block is identified by its numeric disk address( drive 1, cylinder 3, track 2, sector 10)

17

Computer Research Laboratory, EECS, KAIST

File organization module It knows about files and their logical

blocks as well as physical blocks.

It can translate logical block addresses to physical block addresses for the basic file system.

18

Computer Research Laboratory, EECS, KAIST

Logical File System It manages metadata information.

Metadata: It includes all of the file system structure except the actual data.

It manages the directory structure to provide the file organization information with the information.

How?

19

Computer Research Laboratory, EECS, KAIST

Why layered structure?

20

Computer Research Laboratory, EECS, KAIST 21

File System Overview[ On Disk]

Boot block: It contains information needed by the system to boot an operating system.

Volume control block:It contains volume( or partition) details, such as the number of blocks in the partition, size of blocks, free-block count and free-block pointers, free FCB count and FCB pointers.In UNIX, it is called a superblock.

21

Computer Research Laboratory, EECS, KAIST 22

File System Overview A directory structure per file system: In UNIX, it includes file names and

associated inode numbers. A per-file FCB: In UNIX, it is called i-node

22

Computer Research Laboratory, EECS, KAIST 23

In-memory An in-memory mount table:

It contains information about each mounted volume.

An in-memory directory-structure cache:It holds the directory information of recently accessed directories.

System-wide open-file table:It contains a copy of the FCB of each open file.

Per-process open-file table:It contains a pointer to the appropriate entry in the system-wide open-file table.

23

Computer Research Laboratory, EECS, KAIST 24

A Typical File Control Block[ = i-node in Unix]

Computer Research Laboratory, EECS, KAIST 25

In-Memory File System Structures

Computer Research Laboratory, EECS, KAIST 26

Virtual File Systems Virtual File Systems (VFS) provide an

object-oriented way of implementing file systems.

VFS allows the same system call interface (the API) to be used for different types of file systems.

The API is to the VFS interface, rather than any specific type of file system.

Computer Research Laboratory, EECS, KAIST 27

Schematic View of Virtual File System

Computer Research Laboratory, EECS, KAIST 28

Directory Implementation Linear list of file names with pointer

to the data blocks. simple to program time-consuming to execute

Hash Table – linear list with hash data structure. decreases directory search time collisions – situations where two file

names hash to the same location fixed size

Computer Research Laboratory, EECS, KAIST 29

Allocation Methods An allocation method refers to how

disk blocks are allocated for files:

Contiguous allocation

Linked allocation

Indexed allocation

Computer Research Laboratory, EECS, KAIST 30

Contiguous Allocation Each file occupies a set of

contiguous blocks on the disk Simple – only starting location

(block #) and length (number of blocks) are required

Random access Wasteful of space (dynamic

storage-allocation problem) Files cannot grow

Computer Research Laboratory, EECS, KAIST 31

Contiguous Allocation of Disk Space

Computer Research Laboratory, EECS, KAIST 32

Linked Allocation

Computer Research Laboratory, EECS, KAIST 33

File-Allocation Table

Computer Research Laboratory, EECS, KAIST 34

Example of Indexed Allocation

Computer Research Laboratory, EECS, KAIST 35

Combined Scheme: UNIX (4K bytes per block)

Computer Research Laboratory, EECS, KAIST 36

Free-Space Management Bit vector (n blocks)

…

0 1 2 n-1

bit[i] = 0 block[i] free

1 block[i] occupied

Block number calculation

(number of bits per word) *(number of 0-value words) +offset of first 1 bit

Computer Research Laboratory, EECS, KAIST 37

Free-Space Management (Cont.) Bit map requires extra space

Example:block size = 212 bytesdisk size = 230 bytes (1 gigabyte)n = 230/212 = 218 bits (or 32K bytes)

Easy to get contiguous files Linked list (free list)

Cannot get contiguous space easily No waste of space

Grouping Counting

Computer Research Laboratory, EECS, KAIST 38

Free-Space Management (Cont.) Need to protect:

Pointer to free list Bit map

Must be kept on disk Copy in memory and disk may differ Cannot allow for block[i] to have a

situation where bit[i] = 1 in memory and bit[i] = 0 on disk

Solution: Set bit[i] = 1 in disk Allocate block[i] Set bit[i] = 1 in memory

Computer Research Laboratory, EECS, KAIST 39

Directory Implementation Linear list of file names with pointer

to the data blocks simple to program time-consuming to execute

Hash Table – linear list with hash data structure decreases directory search time collisions – situations where two file

names hash to the same location fixed size

Computer Research Laboratory, EECS, KAIST 40

Linked Free Space List on Disk

Computer Research Laboratory, EECS, KAIST 41

Efficiency and Performance Efficiency dependent on:

disk allocation and directory algorithms types of data kept in file’s directory entry

Performance disk cache – separate section of main

memory for frequently used blocks free-behind and read-ahead – techniques

to optimize sequential access improve PC performance by dedicating

section of memory as virtual disk, or RAM disk

Computer Research Laboratory, EECS, KAIST 42

Page Cache A page cache caches pages rather

than disk blocks using virtual memory techniques

Memory-mapped I/O uses a page cache

Routine I/O through the file system uses the buffer (disk) cache

This leads to the following figure

Computer Research Laboratory, EECS, KAIST 43

I/O Without a Unified Buffer Cache

Computer Research Laboratory, EECS, KAIST 44

Unified Buffer Cache A unified buffer cache uses the

same page cache to cache both memory-mapped pages and ordinary file system I/O

Computer Research Laboratory, EECS, KAIST 45

I/O Using a Unified Buffer Cache