Accumulative Versioning File System Moraine and Its Application to Metrics Environment Mame

1
Accumulative Versioning File System Moraine and
Its Application to Metrics Environment Mame

• Tetsuo Yamamoto
• Makoto Matsushita
• Katsuro Inoue ,
• Osaka University, Japan
• Nara Institute of Science and Technology, Japan

2
Disc Capacity of 1995

• 200-------- 0.54Gbytes

Free Space 0.39Gbytes
Linux (Slackware2.2Basic ApplicationsX window)
0.15Gbytes
3
Disc Capacity of 2000

• 200 -------- 40Gbytes

Free Space 39Gbytes
Linux (RedHat7Basic ApplicationsX window)
1Gbytes
4
Benefits of Large Disc Capacity

• Ordinary PC users
• JPEG, MPEG, MP3, HTML, ...
• Our Software Engineers
• ?
• We still use software product management scheme
  which were devised and developed under limited
  disc capacity in old days.

5
Archiving Software Product Evolution

• 5 years ago, it was very expensive, practically
  impossible.
• Now, it may be possible.

FreeBSD 4.1.1

FreeBSD 2.0

23 Release Versions

100Mbytes
270Mbytes
Develop using CVS
1994
2000
Current Archive Size 1Gbytes

• Current disks have enough space to archive
  product evolutions made by each engineer.

6
Development of Moraine and Mame

• Based on this motivation, we have devised
• two things.
• Automatic SE product archive scheme
• -gt Versioning file system Moraine
• Open structured, Non-proprietary
• Quantitative development management framework
• -gt Metrics environment Mame
• (Moraine As a Metrics Environment)

7
Design Policies of Moraine

• Automatic Archiving
• Moraine automatically records the fine-grained
  versions of all files.
• Easy Operation
• Users are not required to learn how to use
  Moraine. To do this, usual file read/write
  operations are hooked, and versioning works are
  performed by Moraine.
• Open Structure
• Moraine does not have to have proprietary data
  repository or versioning tools.

8
Features of Moraine

• Moraine always shows the only latest version in
  the repository to the user.

9
Architecture of Moraine
User
User
Kernel
User Process
VFS
VCD
Vertical File System
Version Control Daemon
Version Management Sub-System
UFS
Unix File system
RCS
VFS is designed as a stackable file system. It is
a logical file system above physical file system
like UFS.
HDD

Control Commands
-retrieve versions -show a delta
10
Stackable File System

• Stackable file system can be constructed without
  changing existent file system.
• The output of a stackable file system is a file
  sent to the lower file system.
• Stackable file system has a portability.

11
Architecture of Moraine
VCD is a daemon process which acts as a bridge
between the kernel and the version management
sub-system.
User
User
Kernel
User Process
VFS
VCD
Vertical File System
Version Control Daemon
Control commands help to manage system behavior.
Version Management Sub-System
UFS
RCS
The version management sub-system performs the
version create work such as delta computing
between versions.
HDD

Control Commands
-retrieve versions -show a delta
12
Evaluation of Moraine

• To know that Moraine is practically acceptable,
  we have measured system performance and stored
  data size.
• Measurement environment
• We have compared Raw UFS (UNIX file system) with
  Moraine.
• Machine Pentium 166MHZ/48MB RAM
• We used such a low performance computer to
  increase comparison sensitivity.

13
Evaluation -Read-

• We have measured a response time for a UNIX
  process to read files.

14
Evaluation -Write-

• We have measured a response time for a UNIX
  process to write files.

15
Evaluation -Write-

• We have measured a response time for a UNIX
  process to write files.

MoraineSync time for completion of all write
operations
16
Evaluation -Data Size- 1/2

• We have applied Moraine to the student project of
  our university.

Student1 Student2 Student3
LOC 9339 4067 2543
of files 45 20 18
Final file size(KB) 225 117 73
of all versions 533 249 357
of versions of source codes 311 147 247
Student1 Student2 Student3
LOC 9339 4067 2543
of files 45 20 18
Final file size(KB) 225 117 73
of all versions 533 249 357
of versions of source codes 311 147 247
17
Evaluation -Data Size- 2/2
18
Overviews of Mame

• Mame uses Moraine to collect metrics data.
• Mame(Moraine As a Metrics Environment) is an
  infrastructure for quantitative product
  measurement of ordinary software development
  activities.

19
Features of Mame

• Engineers do not need to change their working
  environments.
• The data of various activities on the environment
  is automatically collected.
• Mame provides fine-grained data and can be
  abstracted.
• The data format is non-proprietary.

20
Architecture of Mame

• Development environment for software engineers is
  put above Moraine.

Development Activities
Analysis data
Moraine
Read/Write files
VFS
VCD
UFS
RCS
HDD

Commands
21
Experiment of Using Mame

• We have applied Mame to the student project which
  is the same one used as the evaluation of
  Moraine.
• The metrics data was obtained after the project
  termination.
• Metrics data
• The number of files
• The total lines of codes in the files
• The number of C functions in the source files

22
Product Evolution 1/2
147
147

• The horizontal axis is the cumulative number of
  versions for C source files

23
Product Evolution 2/2
147

• The developer first created skeleton of
  functions, then he filled function bodies.

24
Discussions(Moraine)

• Moraine automatically collects all versions of
  all files created or modified.
• We can recover any old files and any old versions
  if necessary.
• The performance of Moraine is enough as a basis
  of software development platforms.

25
Discussions(Mame)

• There are no specific enforcement to the
  developers.
• We can set up or change data collection policy
  during the project or after the project.
• Ordinary metrics environments require
  predetermined data collection policy.

26
Related Work

• COMPAQ(DEC) OpenVMS
• The file system is the proprietary system.
• ClearCase, PVCS, Visual Source Safe
• Users have to recognize the systems and to issue
  check-in/check-out commands for the version
  management.
• 3D Filesystem
• Registering a new version requires some
  supporting tools associated with the file system.

Our Moraine does not need any knowledge or
commands of archiving for users.
27
Conclusion

• We presented Moraine that automatically records
  the history of files. It is very practical.
• We also presented Mame which collects data of
  various activities without a burden to the
  developers.

• This approach would be one answer to rapid
  increase of disk space.

28
The End
29
Version Viewer