A blackboard architecture for control
Artificial Intelligence
SOSP '87 Proceedings of the eleventh ACM Symposium on Operating systems principles
The Hearsay-II Speech-Understanding System: Integrating Knowledge to Resolve Uncertainty
ACM Computing Surveys (CSUR)
Using Expert Systems to Manage Distributed Computer Systems
Using Expert Systems to Manage Distributed Computer Systems
The experimental literature of the internet: an annotated bibliography
ACM SIGCOMM Computer Communication Review
Bibliography on network management
ACM SIGCOMM Computer Communication Review
Efficient use of workstations for passive monitoring of local area networks
SIGCOMM '90 Proceedings of the ACM symposium on Communications architectures & protocols
SIGCOMM '92 Conference proceedings on Communications architectures & protocols
USTC'94 Proceedings of the USENIX Summer 1994 Technical Conference on USENIX Summer 1994 Technical Conference - Volume 1
Computer system performance problem detection using time series models
Usenix-stc'93 Proceedings of the USENIX Summer 1993 Technical Conference on Summer technical conference - Volume 1
Review: Artificial intelligence approaches to network management: recent advances and a survey
Computer Communications
| Hi-index | 0.00 |
Complex, dynamic, and evolving network environments present difficult challenges for monitoring and control. We have encoded some of the expertise of human networking experts into a knowledge-based system that uses production rules and opportunistic scheduling, and have been using this system to better understand the behavior of the TCP/IP protocols and the applications that use them. Novel aspects of this research include understanding how to encode knowledge from this domain, and how to reason efficiently on real networking problems. The prototype system—KNOBS/TCP—can correctly identify many common problems that network experts would normally be required to find (e.g., improper or inefficient retransmission and ACK strategies, silly-window-syndrome, a subset of reset and connection closing anomalies, and basic problems with auxiliary protocols such as address resolution, and routing). Preliminary measurements indicate that the resulting system is reasonably fast and will scale well. In some important test cases, speedup of two orders of magnitude over analogous manual and partially automated techniques has been observed.