GPU-accelerated Evaluation Platform for High Fidelity Networking Modeling

birdsowlΛογισμικό & κατασκευή λογ/κού

2 Δεκ 2013 (πριν από 3 χρόνια και 4 μήνες)

72 εμφανίσεις

GPU
-
accelerated Evaluation
Platform for High Fidelity
Networking Modeling







11 December 2007

Alex Donkers

Joost Schutte

Contents

Summary of the paper


Evaluation


Questions

Summary of the paper

Using commercial graphic cards

to speed up
execution of network simulation models.


Network simulators

high fidelity
performance evaluation


more detailed models

higher computation cost



speed up technique

GPU = graphics processing unit


Computational power GPU against CPU widening.



Computational power of GPU and CPU

(courtesy of Ian Buck, Standford Univ.)

GPU superior because:


Stream processing model


Spatial parallelism


Necessities for GPU usage:


Identification data parallelism in network simultions

Software abstraction


Goal:

Design evaluation platform architecture

Efficient utilisation of computational processors

of GPUs and CPU, memory, IO and other recources.

Available in commodity desktops.


Commodity desktop equipped with multiple GPUs

With Vidia SLI technology more GPUs in singel system.

Suitability for different types of computation:


CPU =

high performance on single thread of execution


GPU =

many more arithmetic units



extremely high

data parallel and





instruction parallel execution


Evaluating process high
-
fidelity network modeling involves:


task
-
parallel computation



multi CPU


data
-
parallel computation


䝐啳


Features necessary for GPU acceleration:


highly data parallel


arithmetic
-
intensive

Power of GPUs is showed by implementing two cases from
a network environment in both CPU and GPU.


Compared are speed and acurracy of the simulation
results.


Two cases:


Fluid
-
flow
-
based TCP model =

predicts the traffic dynamics at





active queue management





routers.


Adaptive antenna model


=

calculates weight of the beam





former in direction minimizing






mean squared error.




Fluid
-
flow
-
based TCP model



TCP flows and active queue management Routers
are modelled with Stochastic differential equations


Transform Stochastic differential equations into
ordinary differential equations (ODEs) for CPU use


CPU
-
based implementation uses a ODE solver,
ODE45, provided in Matlab


GPU maps all data structures in CPU to on
-
board
memory in GPU




Fluid
-
flow
-
based TCP model



Time varying state of routers require
recomputation of ODE solvers periodically


Execution speed of model is significantly
affected by execution speed of ODE solvers


Implementing ODE solver in GPU can
significantly increase size of network that can
be evaluated

Adaptive antenna model




recursively updates weights of the
beamformers in the direction minimizing mean
squared error (MSE)


Recursive least squares (RLS) algorithm is
used


Implement data layout and operations of
arrays of complex numbers in GPU





Evaluation

Strong points

Weak points

Simulation models

Conclusion & Future work

Strong Points

Highly data
-
parallel



Arithmetic
-
intensive

Weak Points

Processes constitute largely sequential
operations

Processes require bit
-
wise operations


Solution: Use DSP platform



Real
-
time simulation

Evaluation simulation models

Hardware Platform:


Dell Dimension desktop


Intel (dual core) 3GHz Pentium 4 CPU


1GB DDR2 memory


nVidia GeForce 7900GTX


512MB texture memory


Vertex & fragment program:



programmed with openGL and GLSL









Simulation models


Differences between GPU & CPU based
simulation for Fluid
-
flow
-
based TCP model



Difference in prediction of traffic dynamics



Difference in execution time


GPU outperforms CPU for with 256 flows & 256
queues or more because of larger number of
iterations in GPU based ODE solver

Normalized ODE solver Evaluation Time

Simulation models

Adaptive antenna model


GPU
-
based simulation runs faster than CPU
-
based one when antenna array size exceeds
256



Execution time of GPU
-
based implementation
linear decreases with respect to the number
of sub
-
carriers due to parallel processing

Simulation Execution Times

Conclusions & Future work

GPU’s can achieve a speedup of 10x
without loss of accuracy

High fidelity network simulations can be
accelerated by parallel use of CPU & GPU
units


Integrate GPU
-
implemented modules into
existing simulation
-
based network
evaluation platform


Questions?