Internet: Announcement

Leader	Don Smith (University of North Carolina, Chapel Hill), smithfd@cs.unc.edu
Meeting	Tuesday 3:00 - 4:00 pm, room 203
Members	Jay Aikat (University of North Carolina, Chapel Hill), aikat@cs.unc.edu Felix Hernandez Campos (University of North Carolina, Chapel Hill), fhernand@cs.unc.edu Steve Marron (SAMSI/University of North Carolina, Chapel Hill), marron@email.unc.edu George Michailidis (University of Michigan), gmichail@umich.edu Cheolwoo Park (SAMSI), cwpark@email.unc.edu David Rolls (SAMSI), rollsd@uncw.edu Haipeng Shen (University of North Carolina, Chapel Hill), haipeng@email.unc.edu
Outline Objectives	Networking research has long relied on simulation as the primary vehicle for demonstrating the effectiveness of proposed algorithms and mechanisms used in routers or TCP/IP protocols. Typically one constructs a network testbed in a laboratory and conducts experiments with actual network hardware and software (or one simulates network hardware and software in software such as the NS network simulator). In either case experimentation proceeds by simulating the use of the (real or simulated) network by a given population of users running applications such as FTP or web browsers. Traffic generators are used to inject application-level data objects into the network according to a model of how the applications or users behave. A critical aspect of this empirical methodology is ensuring that the resulting synthetic traffic, appearing as packets flowing through the network, preserves the essential characteristics of packet flows in real networks. An especially important property to study is the "burstiness" of packet-level traffic because it has been shown to have strong influences on many of the algorithms and mechanisms (e.g., active queue management) that networking researchers study. This working group could consider two important questions: (1) How should we measure and characterize both real and synthetic packet-level traffic so we can verify that synthetic traffic preserves all the essential properties of real traffic? (2) Can we design controlled experiments using a testbed network to confirm various hypotheses and findings from other empirical studies about the physical factors that lead to "burstiness" in traffic?
Webpage	Source-level Trace Replay (If you get a forbidden error, you need to contact Felix Hernandez-Campos, fhernand@cs.unc.edu, to enter this webpage.)

Leader

Don Smith (University of North Carolina, Chapel Hill), smithfd@cs.unc.edu

Meeting

Tuesday 3:00 - 4:00 pm, room 203

Members

Jay Aikat (University of North Carolina, Chapel Hill), aikat@cs.unc.edu
Felix Hernandez Campos (University of North Carolina, Chapel Hill), fhernand@cs.unc.edu
Steve Marron (SAMSI/University of North Carolina, Chapel Hill), marron@email.unc.edu
George Michailidis (University of Michigan), gmichail@umich.edu
Cheolwoo Park (SAMSI), cwpark@email.unc.edu
David Rolls (SAMSI), rollsd@uncw.edu
Haipeng Shen (University of North Carolina, Chapel Hill), haipeng@email.unc.edu

Outline Objectives

Networking research has long relied on simulation as the primary vehicle for demonstrating the effectiveness of proposed algorithms and mechanisms used in routers or TCP/IP protocols. Typically one constructs a network testbed in a laboratory and conducts experiments with actual network hardware and software (or one simulates network hardware and software in software such as the NS network simulator). In either case experimentation proceeds by simulating the use of the (real or simulated) network by a given population of users running applications such as FTP or web browsers. Traffic generators are used to inject application-level data objects into the network according to a model of how the applications or users behave. A critical aspect of this empirical methodology is ensuring that the resulting synthetic traffic, appearing as packets flowing through the network, preserves the essential characteristics of packet flows in real networks. An especially important property to study is the "burstiness" of packet-level traffic because it has been shown to have strong influences on many of the algorithms and mechanisms (e.g., active queue management) that networking researchers study.
This working group could consider two important questions:
(1) How should we measure and characterize both real and synthetic packet-level traffic so we can verify that synthetic traffic preserves all the essential properties of real traffic?
(2) Can we design controlled experiments using a testbed network to confirm various hypotheses and findings from other empirical studies about the physical factors that lead to "burstiness" in traffic?

Webpage

Source-level Trace Replay
(If you get a forbidden error, you need to contact Felix Hernandez-Campos, fhernand@cs.unc.edu, to enter this webpage.)

Meetings:

10/07/2003: Felix Hernandez Campos's presentation

12/16/2003: Queueing Comparison of Original and Replayed Data