NFYK18002U CANCELLED: Concurrent and Distributed Systems

A large set of computer applications are inherently asynchronous, meaning that the program execution order is cannot be scheduled at the time the program is written. The asynchronicity may arise from several sources, such as events, instrument observations, human input, or a combination of program states that trigger a new state. The challenges that stems from asynchronicity are amplified by the fact that any computer CPU has a number of processor-cores, and this number is steadily rising. Thus, working with programs that may have an enormous state-space, and processing these on many-core systems, often triggers errors such as program deadlock, race-conditions, and other more exotic errors. These errors are impossible to test for and hard eliminate with debugging. Thus, the class will introduce a formal algebra, CSP, that provides programmers the means of formally proving that a program is free from a set of known concurrency errors. Tools that automate these proofs and techniques for writing efficient concurrent and correct programs are central to the class. Tools that map solutions to hardware, rather than programs, are also covered.

Many application domains include several, independent, computers that communicate through a network, forming a distributed system. In addition to concurrency errors, such systems must also function in the presence of hardware errors. The class introduces the additional error-models that may arise in distributed systems and the common techniques for addressing these.