Improving Operating System Reliability

Researchers have shown that operating system (OS) 
reliability can be improved by structuring the
 OS as a collection of components protected by 
techniques that limit error propagation. This 
approach is adopted by microkernel systems such 
as Minix3 and L4. When an error occurs in a 
component providing an OS service, the suggested 
recovery strategy is to restart the service in 
order to restore it to a correct state. A restart,
 unfortunately, affects all applications that 
are currently using the service. This is due to
 the loss of state information maintained within 
the service. CuriOS represents a novel OS design 
that uses lightweight partitioning and distribution
 of OS service state to mitigate the problem of 
state loss during a restart. Services are 
encapsulated in separate protection domains 
and access to state information is carefully 
assigned on a need-to-know basis. Error 
propagation within application-related state 
maintained by the service is also significantly 
reduced. This state management approach has
 been implemented for several CuriOS services. 
Experiments show that it is possible to recover
 from 87-100% of all manifested errors while 
maintaining low performance overheads.