Big Talk Storage Study notes (5)

Nine, disk array

1. About LUNs

The LUN is a finer level of address number for the SCSI ID, with more LUN IDs under each SCSI ID. For large disk arrays, hundreds of thousands of virtual disks can be produced, and assigning a SCSI ID to each disk is far from sufficient. Because each SCSI bus allows up to 16 devices to access (currently 32-bit SCSI standard can access 32). The LUN is a secondary addressing ID. A disk array can be virtual multiple LUN addresses under a single SCSI ID, one for each virtual disk. Hardware-level virtual disks, called LUNs, and software-generated virtual disks, called volumes.

2. Dual-Controller disk array

If there is only one controller module inside the array, then it is a spof (single point of Failure), which is the point of failure. To avoid a single point of failure, an additional controller is installed on the disk array, and the original controller shares one or more disk buses on their back end. Two controllers can use Active-standby, or you can use the dual-active of the method.

1) Active-standby

This approach is also called ha mode (high availablity, highly available), that is, two controllers at the same time one is working, the other is waiting, synchronizing and monitoring status. Once the host controller fails, the backup controller takes over the work.

2) dual-active

The implementation of this dual controller is two controllers working simultaneously, each controller on the back-end bus has access, but each bus is usually only one of the controller management, the other controller does not touch. The general number of buses in the back end can be managed by one controller and the other half by another controller. Once one controller is damaged, another controller takes over all of its work.

3. Brain Division

There was a problem with the path between the two controllers at some point, not one of the controllers crashed. Both do not detect each other's presence, and both try to take over all the buses, which simultaneously manipulate all devices.

To prevent this scenario: (1) Use an arbitrator to choose which controller to use to take over all the buses, such as: a disk that can be accessed by one or two controllers, and the controller writes its own quorum information to it. Once a brain division occurs, refer to this disk, who last wrote the information on the control to whom.

(2) with a power supply controller, once one of the controllers to take over, then regardless of whether the other party did not fail or normal, the controller will send a signal to the power supply controller, so that the other side to restart and enter the standby state.

Big Talk Storage Study notes (5)