Date: Mon, 10 Dec 2001 11:35:18 +0200 From: Sheldon Hearn <sheldonh@starjuice.net> To: Kirk McKusick <mckusick@mckusick.com> Cc: freebsd-arch@FreeBSD.ORG Subject: Re: Using a larger block size on large filesystems Message-ID: <53392.1007976918@axl.seasidesoftware.co.za> In-Reply-To: Your message of "Fri, 07 Dec 2001 11:12:57 PST." <200112071913.fB7JCvf29494@beastie.mckusick.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On Fri, 07 Dec 2001 11:12:57 PST, Kirk McKusick wrote:
> FYA, we had this same debate a bit over a decade ago when we changed
> the default from 4K/512 to 8K/1K. Obviously, we decided not to
> special case small filesystems then despite great hand-wringing over
> what would happen...
Thanks. This takes a load off my shoulders with respect to our tendency
to repeat our mistakes in ignorance of our history. :-)
Here's what I intend to commit.
Ciao,
Sheldon.
Index: share/man/man7/tuning.7
===================================================================
RCS file: /home/ncvs/src/share/man/man7/tuning.7,v
retrieving revision 1.29
diff -u -d -r1.29 tuning.7
--- share/man/man7/tuning.7 7 Dec 2001 18:17:37 -0000 1.29
+++ share/man/man7/tuning.7 10 Dec 2001 09:23:48 -0000
@@ -199,19 +199,29 @@
.Pp
.Fx
performs best when using 8K or 16K filesystem block sizes.
-The default filesystem block size is 8K.
-For larger partitions it is usually a good
-idea to use a 16K block size.
-This also requires you to specify a larger
+The default filesystem block size is 16K,
+which provides best performance for most applications,
+with the exception of those that perform random access on large files
+(such as database server software).
+Such applications tend to perform better with a smaller block size,
+although modern disk characteristics are such that the performance
+gain from using a smaller block size may not be worth consideration.
+Using a block size larger than 16K
+can cause fragmentation of the buffer cache and
+lead to lower performance.
+.Pp
+The defaults may be unsuitable
+for a filesystem that requires a very large number of inodes
+or is intended to hold a large number of very small files.
+Such a filesystem should be created with an 8K or 4K block size.
+This also requires you to specify a smaller
fragment size.
We recommend always using a fragment size that is 1/8
the block size (less testing has been done on other fragment size factors).
The
.Xr newfs 8
options for this would be
-.Dq Li "newfs -f 2048 -b 16384 ..." .
-Using a larger block size can cause fragmentation of the buffer cache and
-lead to lower performance.
+.Dq Li "newfs -f 1024 -b 8192 ..." .
.Pp
If a large partition is intended to be used to hold fewer, larger files, such
as a database files, you can increase the
Index: sbin/newfs/newfs.8
===================================================================
RCS file: /home/ncvs/src/sbin/newfs/newfs.8,v
retrieving revision 1.47
diff -u -d -r1.47 newfs.8
--- sbin/newfs/newfs.8 7 Dec 2001 13:18:28 -0000 1.47
+++ sbin/newfs/newfs.8 10 Dec 2001 09:11:49 -0000
@@ -110,7 +110,7 @@
for more details on how to set this option.
.It Fl b Ar block-size
The block size of the file system, in bytes. It must be a power of 2. The
-default size is 8192 bytes, and the smallest allowable size is 4096 bytes.
+default size is 16384 bytes, and the smallest allowable size is 4096 bytes.
The optimal block:fragment ratio is 8:1.
Other ratios are possible, but are not recommended,
and may produce unpredictable results.
@@ -141,7 +141,7 @@
.Ar blocksize Ns /8
and
.Ar blocksize .
-The default is 1024 bytes.
+The default is 2048 bytes.
.It Fl g Ar avgfilesize
The expected average file size for the file system.
.It Fl h Ar avgfpdir
@@ -271,15 +271,19 @@
bad sector allocation.
.El
.Sh EXAMPLES
-.Dl newfs -b 16384 -f 2048 /dev/ad3s1a
+.Dl newfs /dev/ad3s1a
.Pp
Creates a new ufs file system on
.Pa ad3s1a .
.Nm
will use a block size of 16384 bytes, a fragment size of 2048 bytes
and the largest possible number of cylinders per group.
-These values tend to produce better performance than the defaults
-for most applications.
+These values tend to produce better performance for most applications
+than the historical defaults
+(8192 byte block size and 1024 byte fragment size).
+This large fragment size
+may lead to large amounts of wasted space
+on filesystems that contain a large number of small files.
.Sh SEE ALSO
.Xr fdformat 1 ,
.Xr disktab 5 ,
Index: sbin/newfs/newfs.c
===================================================================
RCS file: /home/ncvs/src/sbin/newfs/newfs.c,v
retrieving revision 1.45
diff -u -d -r1.45 newfs.c
--- sbin/newfs/newfs.c 3 Nov 2001 08:35:11 -0000 1.45
+++ sbin/newfs/newfs.c 7 Dec 2001 15:09:18 -0000
@@ -98,8 +98,8 @@
* sectorsize <= DESFRAGSIZE <= DESBLKSIZE
* DESBLKSIZE / DESFRAGSIZE <= 8
*/
-#define DFL_FRAGSIZE 1024
-#define DFL_BLKSIZE 8192
+#define DFL_FRAGSIZE 2048
+#define DFL_BLKSIZE 16384
/*
* Cylinder groups may have up to many cylinders. The actual
Index: usr.sbin/sysinstall/install.c
===================================================================
RCS file: /home/ncvs/src/usr.sbin/sysinstall/install.c,v
retrieving revision 1.312
diff -u -d -r1.312 install.c
--- usr.sbin/sysinstall/install.c 9 Dec 2001 09:47:09 -0000 1.312
+++ usr.sbin/sysinstall/install.c 10 Dec 2001 00:20:21 -0000
@@ -1121,7 +1121,7 @@
variable_set2(SYSTEM_STATE, "update", 0);
else
variable_set2(SYSTEM_STATE, "init", 0);
- variable_set2(VAR_NEWFS_ARGS, "-b 8192 -f 1024", 0);
+ variable_set2(VAR_NEWFS_ARGS, "", 0);
variable_set2(VAR_CONSTERM, "NO", 0);
return DITEM_SUCCESS;
}
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-arch" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?53392.1007976918>