SCALABILITY OF EXT2
Yancan Huang, Guoliang Jin
May 13, 2008
MOTIVATION

Graph for create
MOTIVATION

Graph for open
MOTIVATION

Same method, different graphs:

Code for create:
asmlinkage long sys_creat(const char __user * pathname,
int mode)
{
return sys_open(pathname,
O_CREAT | O_WRONLY | O_TRUNC,
mode);
}

Code for open:
sys_open(pathname, O_RDWR);

Why?

Create: where does the time go?
OVERVIEW
Motivation
 What is scalability of file system
 Experiment environment





Create: where does the time go





Setup
Techniques
Benchmark
The file create process
What does ext2_lookup do
What dose ext2_create do
The file open process
Conclusion & Future work
WHAT IS SCALABILITY

Large File Systems

Large, Sparse Files

Large, Contiguous Files

Large Directories

Large Numbers of Files
OF FILE SYSTEM
EXPERIMENT ENVIRONMENT

Setup: UML
Mount our own ext2 file system called ext2k
 With 1GB empty virtual disk


Measuring techniques
gettimeofday
 long long c;
__asm__ __volatile__ (“rdtsc” : “=A” (c));


Output techniques
printk
 write to log on host


Benchmark

Sequential create and open on a 1GB disk
OVERVIEW
Motivation
 What is scalability of file system
 Experiment environment

Setup
 Techniques
 Benchmark


Create: where does the time go





The file create process
What does ext2_lookup do
What dose ext2_create do
The file open process
Conclusion & Future work
CREATE: WHERE DOES THE TIME GO
asmlinkage long sys_creat(
const char __user * pathname,
int mode)
{
return sys_open(pathname,
O_CREAT | O_WRONLY | O_TRUNC,
mode);
}
THE FILE CREATE PROCESS

The process of sys_open
sys_open {
do_sys_open {
getname
get_unused_fd_flags
do_filp_open {
open_namei
inameidata_to_filp
}
fsnotify_open
fd_install
putname
}
prevent_tail_call
}
// fs/open.c
// fs/open.c
// fs/open.c
// fs/namei.c
THE FILE CREATE PROCESS

When open_namei meets file create
open_namei {
……
path_lookup_create
lookup_hash
open_namei_create
if (! error)
return 0;
……
}
45%
54%
THE FILE CREATE PROCESS

The process of lookup_hash
lookup_hash {
permission
__lookup_hash {
cached_lookup
// always fail for create
struct dentry *new = d_alloc
dentry = inode->i_op->lookup
if (!dentry)
// always true for create
dentry = new
return dentry
}
}
THE FILE CREATE PROCESS

The process of open_namei_create
open_namei_create {
vfs_create {
may_create
security_inode_create
dir->i_op->create
fsnotify_create
}
may_open
}
// dir is an inode
WE ARE NOW IN EXT2
Thanks to inode->i_op->lookup
 Thanks to inode->i_op->create

Going to ext2_lookup
 Going to ext2_create

WHAT DOES EXT2_LOOKUP DO

The process of ext2_lookup
ext2_lookup {
ext2_inode_by_name {
ext2_find_entry {
……
}
}
iget
d_splice_alias
}
WHAT DOES EXT2_FIND_ENTRY DO

The process of ext2_find_entry
ext2_find_entry {
do {
ext2_get_page
ext2_last_byte
while () {
ext2_match
ext2_next_entry
}
} while ()
}
NOW WHERE WE ARE

When open_namei meets file create
open_namei {
……
path_lookup_create
lookup_hash
open_namei_create
if (! error)
return 0;
……
}
WHAT DOES EXT2_CREATE DO

The process of ext2_create
ext2_create {
ext2_new_inode
mark_inode_dirty
ext2_add_nondir {
ext2_add_link
d_instantiate
}
}
WHAT DOES EXT2_ADD_LINK DO

The process of ext2_add_link
ext2_add_link {
for () {
ext2_get_page
ext2_last_byte
while () {
ext2_match
ext2_rec_len_from_disk
}
}
__ext2_write_begin
ext2_commit_chunk
}
HOW OFTEN THESE WHILE LOOP EXECUTED
REVISIT THE CREATE GRAPH
THE FILE OPEN PROCESS

The process of sys_open
sys_open {
do_sys_open {
getname
get_unused_fd_flags
do_filp_open {
open_namei
inameidata_to_filp
}
fsnotify_open
fd_install
putname
}
prevent_tail_call
}
// fs/open.c
// fs/open.c
// fs/open.c
// fs/namei.c
THE FILE OPEN PROCESS

When open_namei meets file open
open_namei {
……
if (!(flag & O_CREAT)) {
error = path_lookup_open(dfd, pathname,
lookup_flags(flag), nd, flag);
if (error)
return error;
goto ok;
}
……
}
THE FILE OPEN PROCESS

The process of path_lookup_open
path_lookup_open {
__path_lookup_intent_open {
get_empty_filp
do_path_lookup {
path_walk {
link_path_walk {
……
}
}
}
}
}
THE FILE OPEN PROCESS

The process of link_path_walk
link_path_walk {
__link_path_open
if (fail) {
dget
mntgrt
__link_path_open
}
}
// in the dcache
// force real lookup requests
REVISIT THE OPEN GRAPH
OVERVIEW
Motivation
 What is scalability of file system
 Experiment environment

Setup
 Techniques
 Benchmark


Create: where does the time go
The file create process
 What does ext2_lookup do
 What dose ext2_create do
 The file open process


Conclusion & Future work
CONCLUSION


In create, ext2_lookup makes sure there won’t be
two files with the same name, and ext2_add_link
performs a similar routine again
The dentry structure of ext2 is linear


Using B-tree to manage this in memory structure
would show better performance
Another scalability issue in ext2 is that its inode
number is determined when the disk is formatted
FUTURE WORK

Try to use B-tree to manage the dentry structure
and test the performance


But the B-tree itself is complex
Try more workload
QUESTIONS?