New features of ZFS on ZOS R13

Noun Explanation:

Shared file system environment: In the parallel Sysplex environment, z/OS UNIX supports all LPARs users to access the entire file system, which is called a shared file system environment.

Owner system (owning system): The Sysplex member of the Mount file system is called the owning system, and the other member is called clients (not owning system).

z/OS Distributed Files Service zSeries file System (ZFS) is the filesystem for z/OS UNIX system services (z/Os USS). The files and directories supported by the ZFS file system can be accessed by z/OS UNIX applications. The ZFS file system, like other local (or remote file systems (HFS,TFS,AUTOMOUNT,NFS), can mount to the tree-like path of z/Os UNIX. ZFS can provide higher performance than previous HFS file systems, and ZFS may completely replace HFS in future z/Os UNIX. z/Os v1r13 ZFS adds some new features compared to previous versions, which are highlighted in the following areas:

1. Enhanced support for the Sysplex sharing mechanism

Starting with z/Os v1r11, ZFS can mount the file system in a Rwshare shared file system environment, but the file system is mount on the owning system and other systems still need to function with the owning system Shipping interaction to access the file system. The z/OS V1r13 enhances support for the Sysplex aware shared file system.

Figure 1-1

As you can see from Figure 1-1, in z/Os V1r13, when a ZFS aggregate mount in a readable and writable manner in a sysplex-aware environment, it automatically mount to all other systems in the Sysplex. Although z/OS USS owning system still exists, the function shipping operations on the USS level between other systems and the owner system have been superseded by more low-level ZFS on R13. Applications on each system will make data requests directly to ZFS in this system, and ZFS determines how to access the file system. The system on which the file system is mount is called the ZFS owner system (ZFS owning systems), which is responsible for updating the metadata of the file to DASD. ZFS on each system can read file content and directory content directly from the file system.

2. Improve access performance for non-owner systems

z/Os V1r13 provides a fully POSIX-compliant asynchronous write-behind operation for a-owner system. Non-owner system can access directories and files directly from disk without having to make an access request to owner system through a ZFS function shipping. Owner system still updates metadata, but the disk space occupied by metadata is very small compared to a complete file content, so the update metadata reduces the overhead on the owner system. This makes owner's status less important, but it improves access performance for non-owner systems. In addition, the loss of interaction between the non-owner system and the owner system greatly reduces CPU overhead on the owner system.

3. Changes in the way files and directories are stored

At the physical IO level, the smallest unit of a ZFS dataset is 4K, but logically ZFS divides the file system into an array of 8K blocks. Each 8K block can be further divided into 8 1k fragment. In previous ZFS, smaller files and directories could be stored in a continuous set of fragments. So multiple files or directories are stored in the same 8 K block. But the practice found that these files or directories are randomly assigned to blocks, so often multiple small files will be placed in different blocks, so does not really save disk space.

In z/Os ZFS, to be able to read or write files or directories directly from any Sysplex member, the fragment is no longer used to store new files or directory data, but the entire 8k block is used to store files or directories. Soft connections and ACLs are still stored in the same 8K block to conserve disk space. On z/Os V1r13 systems, files or directories originally stored in fragment are converted to block format when they are first written (First-write), and content stored throughout the 8K block can be accessed directly by all Sysplex members. Therefore, if the original files and directories are small, the file system space is more tightly used, and after upgrading to Zos V1r13, ZFS may use more disk space because the file storage format needs to be updated, so increasing the space of the file system needs to be considered. But in most cases, space is rarely growing, such as files or directories larger than 8K, or small files that are not placed in the same block. (This depends on the algorithm used on the previous ZFS version)