OS Agents: Using AI Techniques in the Operating System Environment

While recent decades have brought substantial change to the form of the operating system interface, the power of operating system commands has remained nearly constant. Conventional commands, whether visual or textual, specify one particular action to perform. To carry out a complex task, such as reducing disk utilization, the user is forced to explicitly specify each of the necessary steps. Traditional command-language extension mechanisms, such as shell scripts in Unix, enable the user to aggregate and compose various commands, but force him or her to write and debug programs — a formidable challenge for naive users. This paper presents a goal-oriented approach to the operating system command interface, realized through an implementation we call OS agents. Using OS agents, the user simply specifies a goal to accomplish, and the OS agent decides how to accomplish that goal using its knowledge base of the system state and its commands. The OS agent dynamically synthesizes the appropriate command sequence, issues the required commands and system calls, handles errors, and retries commands if necessary. With OS agents, we have applied AI planning and learning techniques to the operating system environment to increase the power of the user’s commands. We have implemented OS agents within a distributed Unix environment. Our experience indicates that it is practical to incorporate novel ideas of automatic planning and learning into contemporary operating systems with a modest amount of work and little performance penalty. In this paper we present OS agents and their operation, and describe the general-purpose mechanism we have provided to flexibly and efficiently support the needs of OS agents within the Unix operating system. This work was supported in part by the National Science Foundation (Grants No. CCR-8907666, CDA-9123308, CCR9200832, and IRI-9211045), Office of Naval Research Grant 92-J-1946, the Washington Technology Center, Digital Equipment Corporation, Boeing Computer Services, Intel Corporation, Hewlett-Packard Corporation, and Apple Computer. Oren Etzioni is supported in part by an NSF National Young Investigator Award. C. Thekkath was supported in part by a fellowship from Intel Corporation. H. Levy was supported in part by a Fulbright research award and by INRIA. Richard Segal is supported by a GTE fellowship.