Large-Scale TCP Models Using Optimistic Parallel Simulation

• Authors: Garrett Yaun;Christopher D. Carothers;Shivkumar Kalyanaraman
• Venue: Proceedings of the seventeenth workshop on Parallel and distributed simulation
• Year: 2003

Internet data traffic is doubling each year, yet bandwidth doesnot appear to be growing as fast as expected and thus shortfalls in available bandwidth, particularly at the "last mile" mayresult. To address these bandwidth allocation and congestionproblems, researchers are proposing new overlay networks thatprovide a high quality of service and a near lossless guarantee.However, the central question raised by these new servicesis what impact will they have in the large? To address these and other network engineering research questions, highperformancesimulation tools are required. However, to date, optimistic techniques have been viewed as operating outside of the performance envelope for Internet protocols, such as TCP, OSPF and BGP.In this paper, we dispel those views and demonstrate thatoptimistic protocols are able to efficiently simulate large-scaleTCP scenarios for realistic, network topologies using a singleHyper-Threaded computing system costing less than $7,000USD. For our real-world topology, we use the core AT&T USnetwork. Our optimistic simulator yields extremely high efficiency and many of our performance runs produce zero rollbacks.Our compact modeling framework reduces the amountof memory required per TCP connection and thus our memoryoverhead per connection for one of our largest experimentalnetwork topologies was 2.6 KB. That value was comprised ofall events used to model TCP packets, TCP connection stateand routing information.