Bob Felderman - Academia.edu (original) (raw)
Uploads
Papers by Bob Felderman
The views and conclusions contained in this document are those of the authors and should not be i... more The views and conclusions contained in this document are those of the authors and should not be interpreted as representing the official
We present our approach for overcoming the cost, oper-ational complexity, and limited scale endem... more We present our approach for overcoming the cost, oper-ational complexity, and limited scale endemic to dat-acenter networks a decade ago. Three themes unify the five generations of datacenter networks detailed in this paper. First, multi-stage Clos topologies built from commodity switch silicon can support cost-effective de-ployment of building-scale networks. Second, much of the general, but complex, decentralized network rout-ing and management protocols supporting arbitrary deployment scenarios were overkill for single-operator, pre-planned datacenter networks. We built a central-ized control mechanism based on a global configura-tion pushed to all datacenter switches. Third, modu-lar hardware design coupled with simple, robust soft-ware allowed our design to also support inter-cluster and wide-area networks. Our datacenter networks run at dozens of sites across the planet, scaling in capacity by 100x over ten years to more than 1Pbps of bisection bandwidth.
Journal of High Speed Networks
Communications of the ACM, 2016
We present two new models and their exact analysis for the problem of two processors running the ... more We present two new models and their exact analysis for the problem of two processors running the Time Warp distributed simulation protocol. Our flrst model addresses the queueing of messages at each processor while the second model adds costs for rollback and state saving. Both models provide insight into the operation of freerunning systems synchronized by rollback.
Workshop on Principles of Advanced and Distributed Simulation, 1992
... BibTeX. @MISC{Cohen92atomic:a, author = {Danny Cohen and ... Multicomputer chips have been us... more ... BibTeX. @MISC{Cohen92atomic:a, author = {Danny Cohen and ... Multicomputer chips have been used at the Information Sciences Institute (USC/ISI) to create a network that bridges the two lowest layers of the networking distance hierarchy: from intra-computer to local area. ...
The views and conclusions contained in this document are those of the authors and should not be i... more The views and conclusions contained in this document are those of the authors and should not be interpreted as representing the official
We present our approach for overcoming the cost, oper-ational complexity, and limited scale endem... more We present our approach for overcoming the cost, oper-ational complexity, and limited scale endemic to dat-acenter networks a decade ago. Three themes unify the five generations of datacenter networks detailed in this paper. First, multi-stage Clos topologies built from commodity switch silicon can support cost-effective de-ployment of building-scale networks. Second, much of the general, but complex, decentralized network rout-ing and management protocols supporting arbitrary deployment scenarios were overkill for single-operator, pre-planned datacenter networks. We built a central-ized control mechanism based on a global configura-tion pushed to all datacenter switches. Third, modu-lar hardware design coupled with simple, robust soft-ware allowed our design to also support inter-cluster and wide-area networks. Our datacenter networks run at dozens of sites across the planet, scaling in capacity by 100x over ten years to more than 1Pbps of bisection bandwidth.
Journal of High Speed Networks
Communications of the ACM, 2016
We present two new models and their exact analysis for the problem of two processors running the ... more We present two new models and their exact analysis for the problem of two processors running the Time Warp distributed simulation protocol. Our flrst model addresses the queueing of messages at each processor while the second model adds costs for rollback and state saving. Both models provide insight into the operation of freerunning systems synchronized by rollback.
Workshop on Principles of Advanced and Distributed Simulation, 1992
... BibTeX. @MISC{Cohen92atomic:a, author = {Danny Cohen and ... Multicomputer chips have been us... more ... BibTeX. @MISC{Cohen92atomic:a, author = {Danny Cohen and ... Multicomputer chips have been used at the Information Sciences Institute (USC/ISI) to create a network that bridges the two lowest layers of the networking distance hierarchy: from intra-computer to local area. ...