D-Link DSN-540 Software User's Guide for DSN-1100-10 - Page 15

Table 1-2. Ways to Organize Volumes Organization JBOD Mirror Stripe Stripe Mirror Parity Definition Redundant One copy of the data is written to the selected Extents. No Two copies of all data are written to independent Yes Extents. Distributes one copy of the data among several drives No to improve the speed of access. Distributes the data among several drives and then Yes keeps a mirror copy of the blocks on each drive Distributes one copy of the data among several drives Yes and adds parity blocks spread throughout the volume to protect against the loss of any single drive. Striped No No Yes Yes Yes Storage Cost 1x 2x 1x 2x x + 1 For organizations that are striped, the data distributed among multiple drives is organized in a series of "stripes". Each stripe consists of a fixed number of blocks on each drive. The fixed number of blocks used on each drive is called a "chunk." This can be viewed as an array where each row is a stripe and each column is a drive. The term "Member" is defined as the number of columns in this array. The xStack Storage system can allocate the required space very flexibly (i.e., an entire member does need to reside on a single drive). Within each member, extents can be obtained from as many drives as necessary. Organizations that are redundant provide protection against loss of data in the event of a failure of any one drive. This adds a storage cost in that some storage will be used for redundancy rather than for customer data. Redundancy requires that the volume data be stored on separate drives, so that data integrity is maintained in the event of a drive failure. For a parity organization, the xStack Storage system distributes the volume into the array described for a stripe organization, and reserves one chunk from each stripe to hold the parity information. The parity information provides redundancy without having to keep a complete copy of the data. If any single drive fails, the parity calculation mechanism can be applied to the data on the remaining drives to provide full access (reading and writing) of all data. Further, the parity calculation mechanism can be used to recreate the data with fresh parity protection onto a new drive to re-gain redundancy. However, this protection comes at a cost: the storage requirement for the volume is increased by one member. For example, if a 10 GB striped volume was distributed among four members, each member would need 2.5 GB of space. To make a parity volume of 10 GB using four members, for instance, each member would need 3.33 GB of space: one chunk in each stripe would be used to hold parity. In other words, to hold the space for parity in a 4-member volume, the xStack Storage system needs to use 1/3 more space than the initiator can access. From another point of view, 1/4 of the space used by that volume is not available to the initiator. The 10 GB volume can also be created using 2.5 GB members if five members are used. As a result, the additional storage cost for a parity volume of four members is 25% (i.e., 25% of the space is not user data, but redundancy data). For a 5-member parity volume, the additional cost is 20%. For a mirror organization, the xStack Storage system divides the volume into two members, each of which contains one full copy of the data. Each member must be allocated on separate drives. The cost of a mirror volume is that data must be stored twice (i.e., for a volume of size x bytes, the xStack Storage system needs 2x bytes, or 100%). For a stripe mirror organization, the xStack Storage system distributes volume into members, as in a striped organization, and then doubles the number of members to allow for the second data copy. As in a mirror organization, the storage cost is double. xStack Storage Management Center Software User's Guide 5