Home
Large file systems require the capacity of several disks, but most file systems must be created on a single device.
A hardware RAID device is one solution to this problem. A hardware RAID device appears as a single device while in fact containing several disk drives internally.
There are other excellent benefits of hardware RAID, but it is an expensive solution if one simply needs to make many small disks look like a single big disk.
Volume managers are the software solution to this problem.
A volume manager is typically a mid-level block device driver (often called a volume driver) which makes many disks appear as a single logical disk In addition to existing in the kernel's block I/O path, a volume manager requires user level programs to configure and manage partitions and volumes.
The virtualized storage perspective produced by volume managers is so useful that often all storage, including hardware RAID, is controlled with a volume manager
A volume manager is a subsystem for on-line disk storage management .
It adds an additional layer between the physical peripherals and the I/O interface in the kernel to present a logical view of disks, unlike current partition schemes where disks are divided into fixed-size sections.
In addition to providing a logical level of management, a volume manager will often implement one or more levels of software RAID to improve performance or reliability.
Veritas Volume Manager
The Veritas Volume Manager (VxVM) is a very advanced product which supports levels of configuration as well as the RAID levels. VxVM runs on HP-UX and Solaris platforms. Java graphical management tools are available in addition to the command line tools.
The abstract objects include: Physical Disks, VM Disks, Subdisks, Plexes, Volumes, and Disk Groups
The VxVM RAID-0 (striping) implementation allows a striped plex to be expanded. This is possible because of the way VxVM virtual objects are constructed. RAID-0 can be used with RAID-1 to combine the advantages of both.
The RAID-1 (mirroring) capabilities can be tuned for optimal performance. Both copies of the data in a mirror are used for reading and adjustable algorithms decide which half of the mirror to read from. Writes to the mirror happen in parallel so the performance slowdown is minimal. There are three methods which VxVM can use for mirror reconstruction depending on the type of failure. Dirty Region Logging is one method which can help the mirror resynchronization process.
The RAID-5 implementation uses logging which prevents data corruption in the case of both a system failure and a disk failure. The write-ahead log should be kept on a solid state disk (SSD) or in NVRAM for performance reasons. Optimizations are also made to improve RAID-5 write performance in general.
The VxVM “Hot-Relocation'' uses hot-spares in software RAID
Large file systems require the capacity of several disks, but most file systems must be created on a single device.
A hardware RAID device is one solution to this problem. A hardware RAID device appears as a single device while in fact containing several disk drives internally.
There are other excellent benefits of hardware RAID, but it is an expensive solution if one simply needs to make many small disks look like a single big disk.
Volume managers are the software solution to this problem.
A volume manager is typically a mid-level block device driver (often called a volume driver) which makes many disks appear as a single logical disk In addition to existing in the kernel's block I/O path, a volume manager requires user level programs to configure and manage partitions and volumes.
The virtualized storage perspective produced by volume managers is so useful that often all storage, including hardware RAID, is controlled with a volume manager
A volume manager is a subsystem for on-line disk storage management .
It adds an additional layer between the physical peripherals and the I/O interface in the kernel to present a logical view of disks, unlike current partition schemes where disks are divided into fixed-size sections.
In addition to providing a logical level of management, a volume manager will often implement one or more levels of software RAID to improve performance or reliability.
Veritas Volume Manager
The Veritas Volume Manager (VxVM) is a very advanced product which supports levels of configuration as well as the RAID levels. VxVM runs on HP-UX and Solaris platforms. Java graphical management tools are available in addition to the command line tools.
The abstract objects include: Physical Disks, VM Disks, Subdisks, Plexes, Volumes, and Disk Groups
The VxVM RAID-0 (striping) implementation allows a striped plex to be expanded. This is possible because of the way VxVM virtual objects are constructed. RAID-0 can be used with RAID-1 to combine the advantages of both.
The RAID-1 (mirroring) capabilities can be tuned for optimal performance. Both copies of the data in a mirror are used for reading and adjustable algorithms decide which half of the mirror to read from. Writes to the mirror happen in parallel so the performance slowdown is minimal. There are three methods which VxVM can use for mirror reconstruction depending on the type of failure. Dirty Region Logging is one method which can help the mirror resynchronization process.
The RAID-5 implementation uses logging which prevents data corruption in the case of both a system failure and a disk failure. The write-ahead log should be kept on a solid state disk (SSD) or in NVRAM for performance reasons. Optimizations are also made to improve RAID-5 write performance in general.
The VxVM “Hot-Relocation'' uses hot-spares in software RAID
No comments:
Post a Comment