ext3 file system

[gusamasan]

Hello, all.



I need evaluate the ext3 file system performance; i need define:

- services provided

- parameters and

- the performance with different parameter values; for example, changin the value of the "data" parameter (journal, ordered, writeback).



I do not know what services ext3 provides. Well, i know intuitively that it provides services to read, write and erase files. But, there are anything more?. Where can i find the API?. Is the ext3 file system POSIX compliant?


Thank you, in advance.

----------------
gusamasan

[aleph]

Is this some kind of assignment from your boss?

- services provided

Do you mean technical support from its maintainers? I guess you can contact Red Hat for that if you're a paid RH customer.

On the other hand, Fedora has no commercial support. It's a distro by the community and for the community.

EDIT:
I re-read your post and saw what you mean. In that case, just read the doc linked below.
tl;dr version: the filesystem supports fsck, online resize, parameter tuning, etc besides the usual read and write.

About the API... what part of API are you interested in? The high-level API such as read(), write(), select() are mostly POSIX-compliant on Linux systems regardless of filesystem. If you're planning to use the filesystem-specific or "internal" API, consult documentation on the Linux kernel. The source code of e2fsprogs may also be of help because it's the no.1 user of ext* internal API.

- parameters

The kernel documentation remains the definitive reference:
http://www.mjmwired.net/kernel/Docum...stems/ext3.txt

- the performance with different parameter values; for example, changin the value of the "data" parameter (journal, ordered, writeback).

data=journal has less chance of corruption in the face of a filesystem failure, but ordered or writeback is faster, and more risky. (http://www.mjmwired.net/kernel/Docum...s/ext3.txt#166). Haven't looked up benchmark results though, but the Web is full of resources.

[forkbomb]

Originally Posted by aleph

Is this some kind of assignment from your boss?

Or professor? My first thought was end-of-semester paper-writing time.

[stevea]

If the is a school assignment - then do your own work.

There is no API for a file system per se. File systems implement various features and these features may or may not be exposed these feature to aplpications via system call APIs.

You will find essentially all of the system calls in section 2 of the manuals. "ls -l /usr/share/man/man2"
Of these many relate specifically to files:
open, close, read, write, lseek - the basic I/O calls.
mmap, mmap32 - more data API
stat (various entry point, fstat, fstat64, ...), chmod ,chown, setfacl ,getfacl - set/get file metadata
link, unlink -

Among other newer features that exist in more recent *nix* file systems are ACLs (part of a POSIX 1002.1e draft) (see getfacl, setfacl). Capabilities - (also part of a POSIX draft) see setfcap, getfcap. Also SELinux context information, see setcontext, getcontext. These later features are built onto the basic file extended attribute data. See "man5 attr" and man 2 {getxattr,setxattr}". THe SELinux extensions never appeared in POSIX.

You'll need to read the POSIX standard IN DETAIL to understand the extent to which the POSIX methods and APIs are fully implemented, partly implemented and implemented in a non-compliant way. For example the default output of "ls -l" is NOT POSIX compliant on Linux, but there is a means to get compliant behavior. If I was looking for non-POSIX behavior in a Linux file system implementation I'd especially check into less used, more complex features like the open(2) O_ASYNC behavior of the flock(2) behavior, the fdatasync(2) behavior. ext3 is probably as POSIX compliant as any file system but that doesn't necessarily imply perfect POSIX compliance.

==

If you want to see the actual file system settable parameters - it's all under "man mke2fs" and check the "also see" list.

[SlowJet]

A professor usually has a PHD and this assignment sounds like a tech person teaching a C.C. class.

SJ

[gusamasan]

Thank you, all.

I am only a student. I am coursing a Performance Evaluation semester.


------------
gusamasan

[stevea]

Originally Posted by gusamasan

Thank you, all.

I am only a student. I am coursing a Performance Evaluation semester.


------------
gusamasan

You will find comparative performance test results for many of the modern Linux file systems are available at:
http://www.phoronix.com

ext3, ext4, btrfs, nilfs2, reiserfs, Reiserfs4 - also comparisons to some *nix FS's like SunMS's zfs.. Just be aware that some of these are still under development and the tests are of a work in progress.

-S

[gusamasan]

Thank you, stevea

I will use those results to compare my measurements.