Deciding on Custom File System Parameters

Before you choose to alter the default file system parameters assigned by the newfs command, you need to understand them. This section describes each of these parameters:

• Block size

• Fragment size

• Minimum free space

• Rotational delay

• Optimization type

• Number of files

Logical Block Size

The logical block size is the size of the blocks that the UNIX kernel uses to read or write files. The logical block size is usually different from the physical block size (usually 512 bytes), which is the size of the smallest block that the disk controller can read or write.

You can specify the logical block size of the file system. After the file system is created, you cannot change this parameter without rebuilding the file system. You can have file systems with different logical block sizes on the same disk.

By default, the logical block size is 8192 bytes (8 Kbytes) for UFS file systems. The UFS file system supports block sizes of 4096 or 8192 bytes (4 or 8 Kbytes). 8 Kbytes is the recommended logical block size.

You can only specify 8192-byte block size on the sun4u platform.

To choose the best logical block size for your system, consider both the performance desired and the available space. For most UFS systems, an 8-Kbyte file system provides the best performance, offering a good balance between disk performance and use of space in primary memory and on disk.

As a general rule, to increase efficiency, use a larger logical block size for file systems where most of the files are very large. Use a smaller logical block size for file systems where most of the files are very small. You can use the quot -c file-system command on a file system to display a complete report on the distribution of files by block size.

Fragment Size

As files are created or expanded, they are allocated disk space in either full logical blocks or portions of logical blocks called fragments. When disk space is needed to hold a data for a file, full blocks are allocated first, and then one or more fragments of a block are allocated for the remainder. For small files, allocation begins with fragments.

The ability to allocate fragments of blocks to files, rather than just whole blocks, saves space by reducing fragmentation of disk space resulting from unused holes in blocks.

You define the fragment size when you create a UFS file system. The default fragment size is 1 Kbyte. Each block can be divided into 1, 2, 4, or 8 fragments, which results in fragment sizes from 8192 bytes to 512 bytes (for 4-Kbyte file systems only). The lower bound is actually tied to the disk sector size, typically 512 bytes.

The upper bound might equal the full block size, in which case the fragment is not a fragment at all. This configuration might be optimal for file systems with very large files when you are more concerned with speed than with space.

When choosing a fragment size, look at the trade-off between time and space: a small fragment size saves space, but requires more time to allocate. As a general rule, to increase storage efficiency, use a larger fragment size for file systems where most of the files are large. Use a smaller fragment size for file systems where most of the files are small.

Minimum Free Space

The minimum free space is the percentage of the total disk space held in reserve when you create the file system. The default reserve is ((64 Mbytes/partition size) * 100), rounded down to the nearest integer and limited between 1% and 10%, inclusively. Free space is important because file access becomes less and less efficient as a file system gets full. As long as there is an adequate amount of free space, UFS file systems operate efficiently. When a file system becomes full, using up the available user space, only root can access the reserved free space.

Commands such as df report the percentage of space that is available to users, excluding the percentage allocated as the minimum free space. When the command reports that more than 100 percent of the disk space in the file system is in use, some of the reserve has been used by root.

If you impose quotas on users, the amount of space available to the users does not include the free space reserve. You can change the value of the minimum free space for an existing file system by using the tunefs command.

Rotational Delay (Gap)

The rotational delay is the expected minimum time (in milliseconds) it takes the CPU to complete a data transfer and initiate a new data transfer on the same disk cylinder. The default delay is zero, as delay-based calculations are not effective when combined with modern on-disk caches.

When writing a file, the UFS allocation routines try to position new blocks on the same disk cylinder as the previous block in the same file. The allocation routines also try to optimally position new blocks within tracks to minimize the disk rotation needed to access them.

To position file blocks so they are "rotationally well-behaved," the allocation routines must know how fast the CPU can service transfers and how long it takes the disk to skip over a block. Using options to the mkfs command, you can indicate how fast the disk rotates and how many disk blocks (sectors) it has per track. The allocation routines use this information to figure out how many milliseconds it takes to skip a disk block. Then using the expected transfer time (rotational delay), the allocation routines can position or place blocks so that the next block is just coming under the disk head when the system is ready to read it.

It is not necessary to specify the rotational delay (-d option to newfs) for some devices.

Place blocks consecutively only if your system is fast enough to read them on the same disk rotation. If the system is too slow, the disk spins past the beginning of the next block in the file and must complete a full rotation before the block can be read, which takes a lot of time. You should try to specify an appropriate value for the gap so that the head is located over the appropriate block when the next disk request occurs.

You can change the value of this parameter for an existing file system by using the tunefs command. The change applies only to subsequent block allocation, not to blocks already allocated.

Optimization Type

The optimization type is either space or time.

• Space - When you select space optimization, disk blocks are allocated to minimize fragmentation and disk use is optimized.

• Time - When you select time optimization, disk blocks are allocated as quickly as possible, with less emphasis on their placement. When there is enough free space, it is relatively easy to allocate disk blocks effectively, without resulting in too much fragmentation. The default is time.

  You can change the value of the optimization type parameter for an existing file system using the tunefs command.

Number of Files

The number of inodes determines the number of files you can have in the file system: one inode for each file. The number of bytes per inode determines the total number of inodes created when the file system is made: the total size of the file system divided by the number of bytes per inode. Once the inodes are allocated, you cannot change the number without recreating the file system.

The default number of bytes per inode is 2048 bytes (2 Kbytes) if the file system is less than one Gbyte. If the file system is larger than one Gbyte, the following formula is used:

If you have a file system with many symbolic links, they can lower the average file size. If your file system is going to have many small files, you can give this parameter a lower value. Note, however, that having too many inodes is much better than running out of them. If you have too few inodes, you could reach the maximum number of files on a disk slice that is practically empty.