IBM i storage management for integrated servers
Integrated servers use virtual storage (virtual disks) that are managed by IBM i.
This brief overview of IBM i storage management concepts is intended for administrators who are more familiar with how x86-based servers manage storage. Some techniques, such as defragmenting, are not necessary in an integrated server environment.
When the integrated server operating system is running, it uses a portion of the IBM i disk capacity. For this reason, the administration of integrated server storage has both an IBM i component and an integrated server operating system component. The IBM i component is used to create and link a chunk of storage to the integrated server. Many of the common disk administration tasks encountered in stand-alone servers (disk drivers, addressing, configuration and protection) are eliminated when you use an integrated server.
Disk storage administration tasks such as formatting and partitioning can be performed on integrated servers in exactly the same way as they are on stand-alone servers.
IBM i and disk drives
The key to understanding how disk storage is allocated to integrated servers is an understanding of how IBM i storage management works. The heart of storage management on IBM i is a technology called single-level storage. Single-level storage is a revolutionary storage management architecture that not only gives IBM i outstanding disk I/O performance, but greatly reduces the amount of administration required. IBM i does not directly manage disk drives. Beneath the operating system a level of software (called Licensed Internal Code) "hides" the disk drives and manages the storage of objects on those disk drives. A virtual address space is mapped over the existing disk space and used for addressing objects rather than disk drive IDs, cylinders, and sectors. Needed objects are copied ("paged in") from this address space on disk into the address space of main memory.
The major features of single-level storage are:
- Single storage pool
The management of physical disk drives is implemented in the Licensed Internal Code, which is similar in concept to the BIOS on a PC.
By default, the operating system and applications see only a single large pool of virtual storage (called the System Auxiliary Storage Pool or system ASP) rather than physical drives. Therefore, the management of physical storage is hidden from the user.
To increase the size of the pool, simply add disk drives to IBM i and they automatically become part of the system ASP. Note that under some circumstances you might create additional storage pools that are called user ASPs and independent ASPs.
- Scattering of data
Instead of an object being stored on a single physical disk drive, single-level storage scatters objects across all physical drives, transparently to the user.
IBM i disk management supports fully parallel disk I/O, which provides outstanding disk I/O performance because each object on the system is accessible by multiple disk arms concurrently.
There is no need to be concerned about particular disk drives filling up, or moving data from one disk to another to improve performance because all data management is taken care of by the licensed internal code. Therefore, IBM i does not require a Database Administrator. Licensed internal code also ensures that there is no disk fragmentation.
- Single address space
Memory and disk on IBM i form a single 64-bit address space.
A single address space enables objects to be accessed by name rather than hardware address, which provides additional integrity and reliability.
Because of the way IBM i manages disk data, you do not generally need to worry about partitioning high-growth databases, defragmenting disks, or disk striping on your integrated server. The integrated server uses device drivers to share the IBM i disk drives. These device drivers send and receive disk data to the IBM i storage management subsystem. IBM i storage management handles the hard disks, including spreading the integrated server disk drive images across multiple hard disk drives and applying RAID and file mirroring (if configured). Disk defragmentation software manages logical file fragmentation of the hard disk images. Because IBM i storage management handles these tasks, running a defragmentation program on the integrated server helps primarily in cases where "critical file system structures" can be defragmented.
Storage pools (ASPs)
In IBM i, physical hard disk drives are pooled together into one storage space called a disk pool, also called an auxiliary storage pool (ASP). If your file system runs out of space, you can add a new hard disk drive to the storage pool, and the new storage space will be available immediately. Every system has at least one storage pool, the system storage pool. The system storage pool is always ASP 1. You can configure additional user storage pools, numbered 2 - 255. You can use storage pools to distribute your IBM i data over different groups of disks. You can also use this concept to move less important applications or data to your older, slower disk drives. Support for independent ASPs (33-255) is provided through IBM Navigator for i. Both the Information Center and IBM Navigator for i refer to ASPs as storage pools or disk pools.
IBM i disks can be protected in these ways:
- RAID-5: The RAID-5 technique groups several disks together to form an array. Each disk holds checksum information of the other disks in the same array. If a disk fails, the RAID-5 disk controller can re-create the data of the failing disk with the help of the checksum information on the other disks. When you replace a failing disk with a new one, IBM i can rebuild the information from the failed disk on the new (and therefore empty) disk.
- Mirroring: Mirroring keeps two copies of data on two different disks. IBM i performs write operations on both disks at the same time, and can simultaneously perform two different read operations on the two disks of a mirrored pair. If one disk fails, IBM i uses information from the second disk. When you replace the failing disk, IBM i copies the data from the intact disk to the new disk.
- Cross-site mirroring: Cross-site mirroring, using the operating system geographic mirroring function for independent ASPs, mirrors data on disks at sites that can be separated by a significant distance.
To further increase the level of protection, you can attach the mirrored disks to two different disk controllers. Then if one controller fails, and with it one set of disks, the other controller can keep the system up. On larger Power server models, you can attach controllers to more than one bus. Attaching the two disk controllers that form a mirrored pair to two different buses increases availability even more.
You can define storage pools on IBM i to have different levels of protection or no protection at all. Then you can put applications and data into a storage pool with the right amount of protection, depending on how important their availability is. For more information about IBM i disk protection and availability options, see the Recovering your system topic collection.