Chapter 7. Disk Cache Basics
I'm going to talk a lot about disk storage and filesystems in this chapter. It is important to make sure you understand the difference between two related things: disk filesystems and Squid's storage schemes.
Filesystems are features of particular operating systems. Almost every Unix variant has an implementation of the Unix File System (UFS). It is also sometimes known as the Berkeley Fast File System (FFS). Linux's default filesystem is called ext2fs. Many operating systems also support newer filesystem technologies. These include names and acronyms such as advfs, xfs, and reiserfs.
Programs (such as Squid) interact with filesystems via a handful of system calls. These are functions such as open( ), close( ), read( ), write( ), stat( ), and unlink( ). The arguments to these system calls are either pathnames (strings) or file descriptors (integers). Filesystem implementation details are hidden from programs. They typically use internal data structures such as inodes, but Squid doesn't know about that.
Squid has a number of different storage schemes. The schemes have different properties and techniques for organizing and accessing cache data on the disk. Most of them use the filesystem interface system calls (e.g., open( ), write( ), etc.).
Squid has five different storage schemes: ufs, aufs, diskd, coss, and null. The first three use the same directory layout, and they are thus interchangeable. coss is an attempt to implement a new filesystem specifically optimized for Squid. null is a minimal implementation of the API: it doesn't actually read or write data to/from the disk.
|
The remainder of this chapter focuses on the squid.conf directives that control the disk cache. This includes replacement policies, object removal, and freshness controls. For the most part, I'll only talk about the default storage scheme: ufs. We'll get to the alternative schemes and other tricks in the next chapter.
