Checkpoint algorithms are generally divided into 2 types: non-coordinated and coordinated.
In the non-coordinated checkpoint algorithm, each process can independently determine the time to save its own state, the disadvantage is that it is easy to produce domino effect mantle. Each process needs to maintain multiple checkpoint files, wasting a lot of work done and lots of storage space.
In the coordination checkpoint algorithm, multiple processes are required to set checkpoints together to achieve a consistent global state, avoid the multi-meter effect, and it only requires each process to maintain a permanent checkpoint in stable storage seven, reducing storage overhead.
In the same-volume recovery fault-tolerant system based on checkpoints and message logs, blocking when accessing solid memory and setting checkpoints is the primary checkpoint overhead, compared to a much smaller communication overhead.
The Intelligent Platform Management interface (Ipmi,intelligent Platform Management Interface) is an industry standard used to manage peripherals used in enterprise-class systems based on Intel architecture, The user can use IPMI to monitor the physical health characteristics of the server, such as temperature, voltage, fan operating status, power state, etc. The standard is customized by companies such as Intel, Hewlett-Packard (Hewlett-Packard), NEC, Dell computers in the United States and Supermicro. The new version is IPMI2.0 (http://www.intel.com/design/servers/ipmi/).