Browse Prior Art Database

A migrating file system based on copy on access

IP.com Disclosure Number: IPCOM000248076D
Publication Date: 2016-Oct-24
Document File: 6 page(s) / 381K

Publishing Venue

The IP.com Prior Art Database

Abstract

This invention introduces a method called Copy On Access(COA) on virtual file system at linux. That is, the necessary file copy is delayed until actually accessing the file. This method works in master/slave manner. It helps both large system and embedded system employment. For deployment on large system, it speeds up the depolyment procedure: allow slave system to run immediately without wait until all files copied from master node; For deployment on embedded system, it automatically generate a minized linux rootfs image, which is meaningful for embedded linux products where disk space is limited. It also saves operator's large effort to determine what files are necessary in the minized linux rootfs image.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 52% of the total text.

Page 01 of 6

A migrating file system based on copy on access

Sometimes, if developer need to customize a linux distribution. The developer wants to generate the most simplified filesystem. Only necessary files are kept.

For example, in a new machine (or docker), The developer want to create a http server, only the httpd related binary files and configure files are needed. How can he remove the unnecessary files?

Normally engineer iterates all files and spot those unnecessary files, and then verify whether the removing of those will lead to any functional problems. However, even shortly it works fine but a seldom code path will easily escape and cause issue later.

Due to the complexity of this procedure, the distribution always includes many unnecessary files.

Known traditional Solution

Drawbacks

Access time of the file system is useful to track the time when the file is written/read.

Firstly, developer can copy all the files into a chroot environment. Then he can set the start time to check. After a long time test, he checks the access time of all files. He can find all the files whose access time is newer than the start time. So he gets all the necessary and dependent files for the test because the access time of all the accessed files have been updated.

No one knows how long the time is enough for testing the function.

Iteratively testing may be needed.

Sometimes, the file system is mounted as no access time. In this case, the files have no access time provided. So the traditional solution doesn't work in this case.

Use nfs i.e. remote file system instead.

Establish a nfs server and setup a nfs file system in the server. Then all the clients could access the remote file system after mounting the remote nfs directory to local filesystem.

Every file access is a remote read/write and it is very inefficient.

Change any files in the master will have impact on client side.

To resolve such issues, a new method will be introduced called Copy On Access(COA). That is, necessary file copy is delayed until accessing the file.

What's the benefit?

1. The most exciting thing is that the system can be put into work immediately´╝îand this solution really fits to the large scale system deployment. One master node controls a plurality of client nodes. This solution can speed up the deployment process.

2. This solution doesn't change the file access time for file attributes' integrity. Since changing the access time might have impact on the system behavior itself. E.g. what if some applications behave differently based on file's access time?

3. The solution...