- 论坛徽章:
- 0
|
Yes, you can use either one or both. The failover mechanism is different between an application and a driver.
How does an application complete its failover? You can image a device with two paths as two houses. The application opens the door to a house, works inside but fails, closes the door, goes to the other house, opens the door, repeats its previous work, and continues next work.
How does a driver finish its failover? You image a device with two paths as two rooms at a house. When the driver fails to work at a room, it walks to the other room to repeats its previous work, and continues next work. If you have more paths (rooms), the driver works around these rooms. So, you can see the failover inside a driver more simple, more quick, more effective.
Also, more than a WRITE SCSI command may be sent to the tape drive when an application asks a write operation for a big data for example to write 3 MB data. When this write operation fails, the application doesn't know which WRITE SCSI command is failed and has to re-write all of 3 MB data. For a driver, it knows the failure occurs on which WRITE command, and just retry this WRITE command. |
|
免费注册
发表于 2012-03-23 15:58
