27-05-2013, 12:24 PM
AMPLE: An Adaptive Traffic Engineering System Based on Virtual Routing Topologies
AMPLE An Adaptive.pdf (Size: 283.63 KB / Downloads: 28)
ABSTRACT
Handling traffic dynamics in order to avoid
network congestion and subsequent service disruptions
is one of the key tasks performed by
contemporary network management systems.
Given the simple but rigid routing and forwarding
functionalities in IP base environments, efficient
resource management and control solutions
against dynamic traffic conditions is still yet to
be obtained. In this article, we introduce
AMPLE — an efficient traffic engineering and
management system that performs adaptive traffic
control by using multiple virtualized routing
topologies. The proposed system consists of two
complementary components: offline link weight
optimization that takes as input the physical network
topology and tries to produce maximum
routing path diversity across multiple virtual
routing topologies for long term operation
through the optimized setting of link weights.
Based on these diverse paths, adaptive traffic
control performs intelligent traffic splitting
across individual routing topologies in reaction
to the monitored network dynamics at short
timescale. According to our evaluation with real
network topologies and traffic traces, the proposed
system is able to cope almost optimally
with unpredicted traffic dynamics and, as such, it
constitutes a new proposal for achieving better
quality of service and overall network performance
in IP networks.
INTRODUCTION
Traffic Engineering (TE) is an essential aspect of
contemporary network management. Offline TE
approaches aim to optimize network resources in
a static manner, but require accurate estimation
of traffic matrices in order to produce optimized
network configurations for long-term operation
(a resource provisioning period each time, typically
in the order of weeks or even longer). However,
these approaches often exhibit operational
inefficiencies due to frequent and significant traffic
dynamics in operational networks. Take the
published traffic traces dataset in the GEANT
network as an illustration. The actual maximum
link utilization (MLU) dynamics is substantial on
a daily basis, varying from less than 40 percent
during off-peak time to more than 90 percent in
busy hours [2]! As such, using one single traffic
matrix as input for offline computing a static TE
configuration is not deemed as an efficient
approach for resource optimization purposes in
such dynamic environments.
SYSTEM OVERVIEW
Figure 2 presents an overall picture of the proposed
AMPLE TE system, with Offline MT-IGP
Link Weight Optimization (OLWO) and Adaptive
Traffic Control (ATC) constituting the key
components. As previously mentioned, the ultimate
objective of OLWO is to provision offline
maximum intra-domain path diversity in the
routing plane, allowing the ATC component to
adjust at short timescale the traffic assignment
across individual VRTs in the forwarding plane.
A salient novelty is that the optimization of the
MT-IGP link weights does not rely on the availability
of the traffic matrix a priori, which plagues
existing offline TE solutions due to the typical
inaccuracy of traffic matrix estimations. Instead,
our offline link weight optimization is only based
on the characteristics of the network itself, i.e.
the physical topology. The computed MT-IGP
link weights are configured in individual routers,
and the corresponding IGP paths within each
VRT are populated in their local routing information
bases (MT-RIBs). While OLWO focuses
on static routing configuration in a long
timescale (e.g. weekly or monthly), the ATC
component provides complementary functionality
to enable short timescale (e.g. hourly) control
in response to the behavior of traffic that cannot
be usually anticipated.
COMPONENT SPECIFICATION
OFFLINE LINK WEIGHT OPTIMIZATION
First of all, a fundamental issue in OLWO is
how to determine the definition of “path diversity”
between PoPs for traffic engineering. Let’s
consider the following two scenarios of MT-IGP
link weight configuration. In the first case, highly
diverse paths (e.g. end-to-end disjoint ones) are
available for some PoP-level S-D pairs, while for
some other pairs individual paths are completely
overlapping with each other across all VRTs. In
the second case, none of the S-D pairs have disjoint
paths, but none of them are completely
overlapping either. Obviously, in the first case if
any “critical” link that is shared by all paths
becomes congested, its load cannot be alleviated
through adjusting traffic splitting ratios at the
associated sources, as their traffic will inevitably
travel through this link no matter which VRT is
used. Hence, our strategy targets the second scenario
by achieving “balanced” path diversity
across all S-D pairs.
NETWORK MONITORING
Network monitoring is responsible for collecting
up-to-date traffic conditions in real-time and
plays an important role for supporting the ATC
operations. AMPLE adopts a hop-by-hop based
monitoring mechanism that is similar to the proposal
of [8]. The basic idea is that a dedicated
monitoring agent deployed at every PoP node is
responsible for monitoring:
• The volume of the traffic originated by the
local customers toward other PoPs (intra-
PoP traffic is ignored).
• The utilization of the directly attached
inter-PoP links.
As shown in Fig. 3, this monitoring agent gathers
data on the locally originated traffic volume
from all the access routers (ARs) attached to
customers at the PoP. Meanwhile the agent also
collects the utilization of the directly attached
inter-PoP links from individual backbone routers
(BRs).
ADAPTIVE TRAFFIC CONTROL
Given the optimized MT-IGP link weights produced
by OLWO, adaptive traffic control (ATC)
can be invoked at short-time intervals during
operation in order to re-optimize the utilization
of network resources in reaction to traffic
dynamics. The optimization objective of ATC is
to minimize the maximum link utilization
(MLU), which is defined as the highest utilization
among all the links in the network. The
rationale behind ATC is to perform periodic and
incremental traffic splitting ratio re-adjustments
across VRTs based on traffic pattern “continuity”
at short a timescale, but without necessarily
performing a global routing re-optimization process
from scratch every time. In this section, we
present a lightweight but efficient algorithm that
can be applied for adaptive adjustment of the
traffic splitting ratio at individual PoP source
nodes to achieve this goal.
WORKING AS A WHOLE SYSTEM
After presenting the detailed information on
individual components, we now briefly describe
how they work in unison as a whole TE system.
First, optimized MT-IGP link weights are configured
on top of the underlying MT-IGP platform
and remain static until the next offline OWLO
cycle. During this period, ATC plays the major
role for adaptively re-balancing the load according
to the traffic dynamics in short-time intervals.
As a bootstrap procedure, the initial traffic splitting
is evenly distributed across VRTs, but this
will be recomputed based on follow-up traffic
monitoring results. In response to the periodic
polling requests by the TE manager, the monitoring
agents attached to individual PoP nodes
report back the incoming traffic volume (from
access routers) and inter-PoP link utilizations
(from backbone routers). The TE manager
accordingly updates the traffic volume between
each S-D pair in the SDPL and link utilization
information stored in the LL of the TIB.
SUMMARY
In this article we have introduced AMPLE, a
novel TE system based on virtualized IGP routing
that enables short timescale traffic control
against unexpected traffic dynamics using multitopology
IGP-based networks. The framework
encompasses two major components, namely,
Offline Link Weight Optimization (OLWO) and
Adaptive Traffic Control (ATC). The OLWO
component takes the physical network topology
as the input and aims to produce maximum IGP
path diversity across multiple routing topologies
through the optimized setting of MT-IGP link
weights. Based on these diverse paths, the ATC
component performs intelligent traffic splitting
adjustments across individual routing topologies
in reaction to the monitored network dynamics
at short timescale. As far as implementation is
concerned, a dedicated traffic engineering manager
is required, having a global view of the
entire network conditions and being responsible
for computing optimized traffic splitting ratios
according to its maintained TE information
base.