Walking the tightrope: responsive yet stable traffic engineering

By Srikanth Kandula, Dina Katabi, Bruce S. Davie, A. Charny. Conference on Applications, Technologies, Architectures, and Protocols for Computer Communication, 2005

472 citations recorded by the source corpus.

Current intra-domain Traffic Engineering (TE) relies on offline methods, which use long term average traffic demands. It cannot react to realtime traffic changes caused by BGP reroutes, diurnal traffic variations, attacks, or flash crowds. Further, current TE deals with network failures by pre-computing alternative routings for a limited set of failures. It may fail to prevent congestion when unanticipated or combination failures occur, even though the network has enough capacity to handle the failure.This paper presents TeXCP, an online distributed TE protocol that balances load in realtime, responding to actual traffic demands and failures. TeXCP uses multiple paths to deliver demands from an ingress to an egress router, adaptively moving traffic from over-utilized to under-utilized paths. These adaptations are carefully designed such that, though done independently by each edge router based on local information, they balance load in the whole network without oscillations. We model TeXCP, prove the stability of the model, and show that it is easy to implement. Our extensive simulations show that, for the same traffic demands, a network using TeXCP supports the same utilization and failure resilience as a network that uses traditional offline TE, but with half or third the capacity.

Subfields: Artificial Intelligence, Computer Networks and Communications

View Paper: https://www.semanticscholar.org/paper/eda902d2b7be0e8c7b5ba76026451e42d75547f7

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Computer Networks and Communications

Citations in the corpus

Achieving high utilization with software-driven WAN (2013, background): “Optimizing WAN e ciency has rich literature including tuning ECMP weights [12], adapting allocations across preestablished tunnels [10, 17], storing and re-routing bulk data at relay nodes [21], caching at application-layer [32] and leveraging reconfigurable optical networks [22].”
GreenTE: Power-aware traffic engineering (2010, methodology): “Since link capacities and trafﬁc matrices are not available for the Rocketfuel topologies, we assign capacities to links using the method described in [16] and generate trafﬁc matrices using the gravity model [17] [18].”
Reducing Network Energy Consumption via Sleeping and Rate-Adaptation (2008, methodology): “In keeping with previous studies [18, 31] we generate trafﬁc as a mix of Pareto ﬂows, and for some results we use constant bit-rate (CBR) trafﬁc.”
MicroTE: fine grained traffic engineering for data centers (2011, background): “Other more responsiveTEproposalsforISP, such as TEXCP[11], rely onlocalloadbalancing decisions which our work shows tobe suboptimalin the context ofDCs.”
Low Latency Geo-distributed Data Analytics (2015, background): “4) Scheduling on the WAN: There has been much work on optimizing WAN transfers including tuning ECMP weights [32] and adapting allocations across pre-established tunnels [31, 39].”
CONGA: distributed congestion-aware load balancing for datacenters (2015, methodology): “CONGA is conceptually similar to TeXCP [26] which dynamically load balances traffic between ingress-egress routers (in a wide-area network) based on path-wise congestion metrics.”
MIRO: multi-path interdomain routing (2006, background): “In addition, recent w ork on TeXCP [16] has explored how to split traffic over multiple i ntradomain paths for more effective traffic engineering.”
HULA: Scalable Load Balancing Using Programmable Data Planes (2016, unknown): “Global utilization-aware load balancing TeXCP [41] and MATE [42] are adaptive trafﬁc-engineering proposals that load balance across multiple ingress-egress paths in a wide-area network based on per-path congestion metrics.”
Layering as Optimization Decomposition: A Mathematical Theory of Network Architectures (2007, methodology): “Various designs have been presented based on somewhat different NUM formulations and decomposition methods: MATE in [34], TeXCP in [59], DATE in [49], Overlay TCP in [46], and others [82], [60], [69], [101], [143].”
Hedera: Dynamic Flow Scheduling for Data Center Networks (2010, background): “TeXCP [23] and MATE [10] perform dynamic traffic engineering across multiple paths in the wide-area by using explicit congestion notification packets, which require as yet unavailable switch support.”
P4p: provider portal for applications (2008, background): “The traditional approach to improving network efficiency is traffic engineering (e.g., [2, 10 ]).”
Augmenting data center networks with multi-gigabit wireless links (2011, methodology): “We use a simple moving average of estimates from the recent past [13].”