SEDA: an architecture for well-conditioned, scalable internet services
SOSP '01 Proceedings of the eighteenth ACM symposium on Operating systems principles
Applied Optimal Control and Estimation
Applied Optimal Control and Estimation
Virtualization Considered Harmful: OS Design Directions for Well-Conditioned Services
HOTOS '01 Proceedings of the Eighth Workshop on Hot Topics in Operating Systems
Failure to thrive: QoS and the culture of operational networking
RIPQoS '03 Proceedings of the ACM SIGCOMM workshop on Revisiting IP QoS: What have we learned, why do we care?
Extended process registry for erlang
ERLANG '07 Proceedings of the 2007 SIGPLAN workshop on ERLANG Workshop
Programming Erlang: Software for a Concurrent World
Programming Erlang: Software for a Concurrent World
Validity of the single processor approach to achieving large scale computing capabilities
AFIPS '67 (Spring) Proceedings of the April 18-20, 1967, spring joint computer conference
Software agents mobility using process migration mechanism in distributed Erlang
Proceedings of the twelfth ACM SIGPLAN workshop on Erlang
| Hi-index | 0.01 |
Although Telecoms, the domain for which Erlang was conceived, has strong and ubiquitous requirements on overload protection, the Erlang/OTP platform offers no unified approach to addressing the problem. The Erlang community mailing list frequently sports discussions on how to avoid overload situations in individual components and processes, indicating that such an approach would be welcome. As Telecoms migrated from carefully regulated single-service networks towards multimedia services on top of best-effort multi-service packet data backbones, much was learned about providing end-to-end quality of service with a network of loosely coupled components, with only basic means of prioritization and flow control. This paper explores the similarity of such networks with typical Erlang-based message-passing architectures, and argues that a robust way of managing high-load conditions is to regulate at the input edges of the system, and sampling known internal choke points in order to dynamically maintain optimum throughput. A selection of typical overload conditions are discussed, and a new load regulation framework - JOBS - is presented, together with examples of how such overload conditions can be mitigated.