|
|
|
|
|
When creating a volume in Veritas Volume Manager in previous releases, you could specify the disk storage on which to lay out its various parts, subdisks, plexes, and so on. In specifying the storage to be used, you had to take into account the tolerance of a volume to failure of any component of the storage infrastructure, and how the specified layout affected I/O performance and reliability of service. For small installations with a few tens of disks in relatively low-specification arrays, you could either specify the storage layout manually to commands such as vxassist, or rely on vxassist to choose appropriate storage based on general layout specification, such as "mirror across controllers" and "mirror across enclosures," and using the set of heuristic rules that were hard-coded within vxassist.
Although some storage attributes are known to VxVM, you must do most of the work in deciding how to lay out the storage if you are to create a volume with the desired performance, reliability and fault tolerance.
Figure 1-1 shows the traditional model for allocating storage to volumes.
Traditional model for creating and administering volumes in Veritas Volume Manager
Click the thumbnail above to view full-sized image.
When intelligent disk arrays are used, many sophisticated features, such as RAID capabilities, snapshot facilities, and remote replication, are provided by logical unit storage devices, LUNs, that are exported by the disk array. Such devices may or may not have ways of making their attributes known to VxVM. In any case, you may be presented with hundreds or thousands of LUNs connected over a SAN.
Allocating storage to volumes when faced with a potentially large number of devices with widely varying and possibly hidden properties is a daunting task to perform manually. ISP aids you in managing large sets of storage by providing an allocation engine that chooses which storage to use based on the capabilities that you specify for the volumes to be created.
The main differences between the traditional model for creating volumes and ISP are that the set of information about the available storage is potentially unlimited, and the set of rules that the allocation engine uses to choose storage is defined externally to commands such as vxassist and vxvoladm.
Figure 1-2 illustrates how ISP improves on the traditional model for creating volumes.
How ISP enhances volume management
Click the thumbnail above to view full-sized image.