[Ans-research] SNS release v0.1 - Scalable Network Simulator

["Emin Gun Sirer" <egs+sns at cs.cornell.edu>]

Hi,

We are writing to announce the public release of SNS v0.1, a
Staged Network Simulator, designed for efficient simulation of
large-scale wireless networks.

SNS derives its performance and scale through "staged simulation" [1,2].
Staged simulation is a new technique for reducing redundant
computations commonly encountered in wireless network simulations.
Simulations often exhibit redundancy, both within a single run
as well as across a batch of runs with similar scenarios. The
core behind staging is to compute expensive functions once,
cache the results, and reuse them whenever possible. Since most
computations in a traditional wireless simulator are directly or
indirectly dependent on simulation time and provide little opportunity
for reuse, staging proposes accuracy-preserving transformations to
restructure the internal events in a simulator to break this
dependency & expose and eliminate the redundant computations. These
transformations include currying, incremental-computation, computing
auxiliary results and time-shifting, and are documented in our papers.
The end result is a simulator where basic, expensive operations, such 
as neighborhood computations, can be cached and reused without having
to be computed from scratch.

SNS is a branch off of the ns2 code base (version 2.1b9 with CMU
Monarch extensions), where we replaced the guts of the simulator, 
including the event scheduler, to perform staging while we preserved
the public ns2 interface. As a result, legacy code written for ns2 in 
C++ and Tcl continues to work on SNS, though faster and at higher
scales. SNS can store the results of past computations on disk and reuse
them across different runs of the simulator - a feature we found to be
very useful when performing simulations in batches.

Staging yields significant gains in both simulation performance and
scale. Whereas stock ns2, under commonly-used ad hoc networking setups
with N nodes, scales as O(N^2), SNS's execution time scales with O(N).
On our test platform (1.8 GHz Intel Xeon with 6 GB RAM), this
translates to a ~50-fold speedup for a 1500 node network, where the 
execution time improves from 32 hours to 37 minutes.

SNS is free and can be downloaded from:
     http://www.cs.cornell.edu/People/egs/sns/

Hope this is a useful tool for your research. When referring to SNS,
please cite our TOMACS paper instead of the earlier WSC paper.

Gun & Kevin.

--

[1] Kevin Walsh, Emin Gun Sirer. Staged Simulation: A General Technique
for Improving Simulation Scale and Performance. To appear in ACM
Transactions on Modeling and Computer Simulation (TOMACS), April 2004.

[2] Kevin Walsh, Emin Gun Sirer. Staged Simulation for Improving the Scale
and Performance of Wireless Network Simulations. In Procedings of the
Winter Simulation Conference, New Orleans, LA, December 2003.


_______________________________________________
Ans-research mailing list
Ans-research@ietf.org
https://www1.ietf.org/mailman/listinfo/ans-research