|
Each logical volume in the IBM XIV system is divided into multiple stripes of 1MB. These stripes are spread over all the disks in the system, using a sophisticated pseudo-random distribution mechanism. The XIV system's revolutionary approach ensures that all disks and modules are used equally, regardless of access patterns. The system keeps the load perfectly balanced, even if certain applications access certain volumes, or certain parts within a volume, more frequently. Distributed Rebuild and Reduced Overhead
Current methods for implementing storage redundancy involve mirrored pairs of disks or RAID-5 disk groups. In most other systems, rebuild time takes between 6 and 25 hours, depending on disk size and protection scheme. The disk I/O-consuming rebuild process typically causes severe performance degradation until the rebuild is over. The IBM XIV system's revolutionary redundancy scheme provides a distributed rebuild mechanism in which all disks participate in a given rebuild, with each disk rebuilding only a small portion. As a result, rebuild process overhead is minimal, and performance levels are unaffected. Rebuild time with IBM XIV is as little as 30 minutes (or less!) for a 1 TB drive.
No Write-through Mode
To achieve redundancy, all storage architectures must store written data in two cache units before the host is acknowledged. Traditional systems, due to their architecture, are unable to maintain cache redundancy upon failure of a cache unit or UPS. These systems switch to "write-through" mode, acknowledging the host only after the write transaction has been committed to the disks. This in turn reduces the performance level during work so much as to be practically the equivalent of shutting down the system.
The XIV system never uses write-through mode. Its flexible new architecture enables write requests to be cached in two different modules, even after module or UPS failure.
|
