Abstract— path for each incoming packet. On the basis

Abstract— A well designed selection
strategy is a crucial aspect in determining network performance of an adaptive
router under dynamic traffic conditions. This paper presents a review of
research on selection strategies, differentiates the prevalent schemes into
three categories: local adaptive selection, composite regional selection and
rule based selection. We compared the selection strategies according to metric
chosen, selection method, routing algorithm used, traffic models and
constraints involved. The study shows that the efficiency of an adaptive router
is governed by metric chosen for congestion measurement and its impact on
network performance. An optimal selection strategy can be
designed by considering the network status from both local as well as
non-local parameters. Further use of fuzzy controller for selection function
design can improve the performance by removing the ambiguities in congestion
measurement for the accepted directions.

 

Keywords— Selection function, congestion, adaptive routing, network on chips,
local metrics, global metrics,

We Will Write a Custom Essay Specifically
For You For Only $13.90/page!


order now

 

 INTRODUCTION

       With
an increase in complexity and scaling of design, network-on-chips (NOC)
emerged  as fast and low power way of
communication for future systems-on-chip (SOC). Many properties like network
topology, routing policy, flow control, switching mechanism, etc. determine the
overall performance of a network. One such aspect is routing algorithm  implementation by selecting a suitable and a deadlock
free output path for each  incoming
packet. On the basis of path selection, routing algorithm is classified as
deterministic, oblivious and adaptive routing. The path  from the source node to destination node is governed
by the source and the destination itself and  network traffic status is not considered in
deterministic routing. In oblivious routing, a packet can follow multiple paths
from a source router to a destination , but the selection of path is
independent of network status. In an adaptive routing, each incoming packet is
forwarded to a destination node after considering the  network status of the path 1. The
deterministic routing algorithms despite being simple and less complex are not
able to maintain the evenness across the links under non-uniform or bursty
traffic. On the other side as adaptive routing algorithm uses network status
for selecting a path, chances of routing through a congested or faulty region
reduces considerably. An adaptive routing algorithms is divided into two
functions: routing function and selection function as shown in figure 1. The
routing function decides the possible set of output channels based on the
source  and the destination nodes.  An output port  is selected from  these possible channels using selection
function based on the status of output channels at the current node. This
selection should be such so as to supply available free channel. In any case of
availability of more than one output channel, a strategy is required to select
one of them 

x

Hi!
I'm Dianna!

Would you like to get a custom essay? How about receiving a customized one?

Check it out