Research And Implementation On Device Hot-Plug Technologies In Server Operating System

Device fault is one of the most substantial causes for server crash. Great financial lost may be encountered due to the breakdown or long-time outage of key server. Under the pressure of the cost reduction, the thorough overwhelming of device fault is not more than a dream. To ensure the availability of the system, effective fault recovery method is essential. The methods used as well as the time needed in the recovery procedure of heavy loaded devices often become the key to the availability of the whole system.Hot-Plug provides a way to add and remove devices without disturbing the normal operation of the server. Serving as a repair mean, Hot-Plug can avoid the breakdown or long-time outage caused by the cold insertion or removal. Hot-Plug itself starts as a hardware mechanism and end up as a software technology. It is our goal to build a maintenance-oriented mechanism to solve device fault using Hot-Plug inside the server operating system.First, we make a brief introduction to the Hot-Plug technology and related topics. According to the common usage model of Hot-Plug and the characteristic of the OS, we design and implement KHPSA: the KYLIN Hot-Plug Software Architecture.Second, with the KYLIN device driver framework extended, we design and implement the KPEHK: the KYLIN PCI Express native Hot-Plug Kernel. Inside KPEHK, we put forward a PCI Bridge Resource Rebalance algorithm.Finally, under representative condition, we analyze the impact of the Hot-Plug ability brought by KPEHK to the characteristic of server system reliability .We also make a discussion on the preferred cycle for the administrative Hot-Plug of KPEHK when cost factor is concerned.Analysis shows that the maintenance based on Hot-Plug enhances KYLIN's capability in preserving the availability of the server system.