DualFS: a New Journaling Journaling File System without File System - - PowerPoint PPT Presentation

dualfs a new journaling journaling file system without
SMART_READER_LITE
LIVE PREVIEW

DualFS: a New Journaling Journaling File System without File System - - PowerPoint PPT Presentation

International Conference on Supercomputing International Conference on Supercomputing (ICS 2002) (ICS 2002) New York, New York, June 2002 New York, New York, June 2002 DualFS: a New Journaling Journaling File System without File System


slide-1
SLIDE 1

GACOP D A C U C P

International Conference on Supercomputing International Conference on Supercomputing (ICS 2002) (ICS 2002)

New York, New York, June 2002 New York, New York, June 2002

DualFS: a New Journaling File System without Meta-Data Duplication DualFS: a New DualFS: a New Journaling Journaling File System without File System without Meta Meta-

  • Data Duplication

Data Duplication

Juan Piernas1, Toni Cortes2, and José M. García1

1 Dpt. Ingeniería y Tecnología de Computadores

Universidad de Murcia

2 Dpt. d’Arquitectura de Computadors

Universitat Politècnica de Catalunya e-mail: piernas@ditec.um.es

slide-2
SLIDE 2

GACOP D A C U C P

Introduction Introduction Introduction

Meta-data management is a key design issue

Especially important for recovery after a system crash

Traditional file systems:

Write meta-data in a synchronous way Use fsck-like tools

New approaches:

Log of last meta-data updates (e.g. XFS, JFS) Asynchronous meta-data writes (e.g. Soft Updates)

New approaches treat data and meta-data somewhat differently

But they are completely different.

slide-3
SLIDE 3

GACOP D A C U C P

Objective Objective Objective

Introduce a new high performance journaling file system, DualFS Present a working prototype Evaluate the prototype and ideas

slide-4
SLIDE 4

GACOP D A C U C P

Outline Outline Outline

Introduction DualFS Experimental Methodology and Results Conclusions

slide-5
SLIDE 5

GACOP D A C U C P

Key Issues Key Issues Key Issues

Main idea:

Manage data and meta-data in completely different ways. Each type of blocks will be located on different devices:

– meta-data blocks in the meta-data device – data blocks in the data device – can share the same disk

Previous work

Muller and Pasquale (SOSP’91) Anderson, Chase, and Vahdat. (OSDI’00)

slide-6
SLIDE 6

GACOP D A C U C P

Data Device Data Device Data Device

Like Ext2 without meta-data blocks

Standard linux file system

Structure based on the concept of group of data blocks

Related blocks are kept together. Grouping is performed in a per directory basis.

Data blocks are not written synchronously

slide-7
SLIDE 7

GACOP D A C U C P

Meta-Data Device Meta Meta-

  • Data

Data Device Device

We understand meta-data as all these items:

i-nodes indirect blocks directory “data” blocks symbolic links bitmaps superblock copies

Organized as a log-structured file system

Similar to BSD-LFS. Similar cleaner

slide-8
SLIDE 8

GACOP D A C U C P

Meta-Data Device Structure Meta Meta-

  • Data

Data Device Structure Device Structure

slide-9
SLIDE 9

GACOP D A C U C P

IFile IFile IFile

slide-10
SLIDE 10

GACOP D A C U C P

Recovery Recovery Recovery

DualFS is considered consistent when information about meta-data is correct. We can recover the file system consistency very quickly from the last checkpoint. The length of time for recovery is proportional to the inter-checkpoint interval.

slide-11
SLIDE 11

GACOP D A C U C P

Potential and limitations Potential and limitations Potential and limitations

Advantages of DualFS

One-copy meta-data elements

– Traditional journaling file systems write meta-data twice

In the log In the regular file system structure

Different structures for the meta-data and the data

devices

Potential Problem:

Read-only workloads

– Possible solution: to put meta-data blocks on another disk

slide-12
SLIDE 12

GACOP D A C U C P

Outline Outline Outline

Introduction DualFS Experimental Methodology and Results Conclusions

slide-13
SLIDE 13

GACOP D A C U C P

Experimental Methodology Experimental Methodology Experimental Methodology

DualFS has been compared with:

Ext2, the default file system in Linux, and Ext3, a journaling file system derived from Ext2.

Tested applications

Microbenchmarks (in the paper) Macrobenchmarks

slide-14
SLIDE 14

GACOP D A C U C P

System Under Test System Under System Under Test Test

Linux Platform Processor Two 450 Mhz Pentium III Memory 256 MB, PC100 SDRAM Disk Two 4 GB IDE 5,400 RPM Seagate ST- 34310A. Two 4GB SCSI 10,000 RPM FUJITSU MAC3045SC. SCSI disk 1: Operating system, swap and trace log. SCSI disk 2: trace log. IDE disks: test disks OS Linux 2.2.19

slide-15
SLIDE 15

GACOP D A C U C P

Macrobenchmarks Macrobenchmarks Macrobenchmarks

Kernel Compilation for 1 Process (KC-1P)

make dep, make bzImage, make modules

Kernel Compilation for 8 Processes (KC-8P)

make dep, make -j8 bzImage, make -j8 modules

Video Compression (VC) Specweb99 (SW99) PostMark (PM)

slide-16
SLIDE 16

GACOP D A C U C P

Macrobenchmarks Results Macrobenchmarks Results Macrobenchmarks Results

Macrobenchmarks (1 disk)

982.19 sec 309.06 sec 62.36 sec 84.87 sec 51.65 sec 1.11 2.42 1.78 1.14 1.26 0.57 1.01 0.63 0.56 0.24

0.5 1 1.5 2 2.5 3 KC-1P KC-8P VC SW99 PM

Benchmark Normalized Disk I/O Time

Ext2 Ext3 DualFS

slide-17
SLIDE 17

GACOP D A C U C P

Cleaner Evaluation Cleaner Evaluation Cleaner Evaluation

Experiment for evaluating the cleaner :

copy a directory tree delete 87.5% (7/8) of its regular files Repeat the two previous steps 20 times

Tested DualFS configurations:

DualFS without cleaner DualFS with cleaner

– cleaning a segment every five seconds.

slide-18
SLIDE 18

GACOP D A C U C P

Cleaner Results Cleaner Results Cleaner Results

44.9 64.04 42.16 44.41

10 20 30 40 50 60 70 Ext2 Ext3 DualFS-cleaner DualFS+cleaner

File System I/O Time

slide-19
SLIDE 19

GACOP D A C U C P

Outline Outline Outline

Introduction DualFS Experimental Methodology and Results Conclusions

slide-20
SLIDE 20

GACOP D A C U C P

Conclusions Conclusions Conclusions

DualFS is a new journaling file system

data and meta-data managed in very different ways

  • ne-copy meta-data blocks

We have compared DualFS against Ext2 and Ext3:

DualFS reduces total I/O time in most cases (up to 97%) DualFS increases total I/O time (up to 17%)

– Only in a few and limited cases

DualFS has a very good performance in

– Internet – Development environments

slide-21
SLIDE 21

GACOP D A C U C P

International Conference on Supercomputing International Conference on Supercomputing (ICS 2002) (ICS 2002)

New York, New York, June 2002 New York, New York, June 2002

DualFS: a New Journaling File System without Meta-Data Duplication DualFS: a New DualFS: a New Journaling Journaling File System without File System without Meta Meta-

  • Data Duplication

Data Duplication

Juan Piernas1, Toni Cortes2, and José M. García1

1 Dpt. Ingeniería y Tecnología de Computadores

Universidad de Murcia

2 Dpt. d’Arquitectura de Computadors

Universitat Politècnica de Catalunya e-mail: piernas@ditec.um.es