IBM SAN Volume Controller

The SAN Volume Controller (SVC) is in-band virtualization and combines disk subsystems from multiple vendors into a single reservoir of capacity that can be managed from a single point. While the initial disk configuration and presentation must be made using the utility unique to each type of subsystem, once defined all the disks are managed from a central SVC console.

The SVC can enable a tiered storage environment by physically connecting the different classes of storage devices together, and configuring them as storage pools with different availability and cost characteristics. The SVC does not provide an automated means to migrate data between the pools.
The image below shows the principal components of the SVC. Managed Disks (MDisks) correspond to LUNs on the disk subsystems, and are combined together into Managed Disk Groups (MDG). Generally each MDG will have different performance and price characteristics. The MDisks are split up into extents, and the extents are combined together to form Virtual Disks (VDisks). An extent is a fixed size block of data, which you specify when you define the MDG. Extents can vary in size from 8k to 512k, with a default of 64k. You pick the extent size that best suits your application needs. The VDisks are then presented to the hosts as SCSI or FC disks. VDisks can be full size or they can be space efficient. You can also mirror a VDisk inside the SVC I/O group for resilience.
An SVC cluster is a collection of SVC nodes. These are grouped into pairs called I/O groups. A cluster contains between 1 and 4 I/O groups and so between 2 and 8 SVC nodes. Each I/O group owns one or more MDGs and their respective VDisks Each IO group can connect to up to 256 hosts.
SAN zoning is critical to the operation of an SVC Cluster and while the diagram below does not show the SAN switches that are used to connect the SVC cluster, it does show the SAN zoning in pink. The physical disks must be kept in a separate zone from the hosts, and each host should be it its own zone with its VDisks You may also want to define a SAN zone for the SVC nodes in the cluster to allow them to communicate together.

Typically, an SVC install will comprise the following steps:

• Sort out the SAN Zoning, ideally as two separate fabrics for resilience.
• Create the SVC Cluster. Initially this must be done from the hardware console on a single SVC node. You will need the IP address of the cluster to do this.
• Complete the cluster definition using the SSPC then add the other SVC nodes to the cluster SSPC.
• Check and enable your SVC licensing
• Create Manage Disk Groups
• Format your external disk subsystems and install HBA drivers and multi-path drivers.
• Add MDisks to the MDGS.
• Create VDisks
• Create Host objects to correspond to the hosts that you plan at attach the VDisks to. You can define the Host Objects as using either SCSI or FC connectivity.
• Map VDisks to Hosts

The SVC can support advanced copy services between unlike devices, either high-cost to low-cost or between different vendors. The copy services facilities are FlashCopy and Metro Mirror, referring to copy within an SVC cluster and copies between remote SVC clusters. The smallest unit of granularity for Flashcopy is the virtual disk. It is possible to flash data between different MDGs and so to different classes of disk, but the virtual disks must be identically sized. Metro Mirror is analogous to PRRC remote copy and supports distances up to 10km apart. Again it is possible to mirror data between different classes of disk, but the virtual disks must be identically sized.
IBM introduced Global Mirror, the asynchronous long distance replication product, with SVC version 4.1, in May 2006. SVC 4.1 also has support for 4 Gb/s connectivity, as well support for extra operating systems and other manufactures storage devices.

The SVC can insulate host applications from changes to physical storage. This means that physical storage can be added or removed transparently. The SVC includes a dynamic data migration facility to move data between physical volumes and multi-pathing software for failover and load balancing. It is possible to move data around inside the SVC by VDisk, MDisk or at extent level. If you want to migrate data onto the SVC, define a VDisks of the same size as the LUN you want to copy in image mode, then just associate the VDisk with that LUN. The data is now part of the SVC and that is the only way to attach external disks to the SVC without destroying the data.

The SVC is hosted on xSeries servers or the Cisco MDS 9000 family of switches. An I/O group is a pair of servers and up to four I/O groups can be clustered for resilience and bandwidth. Each cluster of two to eight servers supports up to 4096 virtual volumes. It incorporates an SMIS standard API and supports LAN free and Server free backups.

back to top

Copyright © Lascon Storage Ltd. 2000 to present date. By entering and using this site, you accept the conditions and limitations of use