Efficient and Cost-Effective Hybrid Congestion Control for HPC Interconnection Networks

Escudero-Sahuquillo, Jesús; Gran, Ernst Gunnar; García, Pedro J.; Flich, José; Skeie, Tor; Lysne, Olav; Quiles, Francisco J.; Duato, J.

doi:10.1109/tpds.2014.2307851

Cited by 23 publications

(8 citation statements)

References 15 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Fig. 1a shows the runtime for MILC and GPCNeT under low (0.25% percent time stalled 3 ) and high (7.5% percent time stalled) congestion, as measured from network stall counters. The figure shows that response to congestion in the system (that is captured through network counters) varies.…”

Section: Motivation Applications Running In Hpc Systemsmentioning

confidence: 99%

Application-aware Congestion Mitigation for High-Performance Computing Systems

Patke,

Jha,

Qiu

et al. 2020

Preprint

View full text Add to dashboard Cite

High-performance computing (HPC) systems frequently experience congestion leading to significant application performance variation. However, the impact of congestion on application runtime differs from application to application depending on their network characteristics (such as bandwidth and latency requirements). We leverage this insight to develop Netscope, an automated ML-driven framework that considers those network characteristics to dynamically mitigate congestion. We evaluate Netscope on four Cray Aries systems, including a production supercomputer on real scientific applications. Netscope has a lower training costs and accurately estimates the impact of congestion on application runtime with a correlation between 0.7 and 0.9 for common scientific applications. Moreover, we find that Netscope reduces tail runtime variability by up to 14.9× while improving median system utility by 12%.

show abstract

Section: Motivation Applications Running In Hpc Systemsmentioning

confidence: 99%

Application-aware Congestion Mitigation for High-Performance Computing Systems

Patke,

Jha,

Qiu

et al. 2020

Preprint

View full text Add to dashboard Cite

show abstract

“…Although injection throttling does not directly eliminate HoL blocking, this effect disappears once congestion is removed. Moreover, injection throttling is not scalable with network size as the delay from congestion detection to reaction at the sources grows with the network diameter [14]. By contrast, queue-based flow-separation strategies are based on having at each port a set of queues (i.e., Virtual Channels (VCs) [9]) to store separately different packet flows, preventing or reducing the impact of HoL blocking.…”

Section: Solutions To Hol Blockingmentioning

confidence: 99%

“…Some of these solutions, such as the one described for ATLAS [22], identify hot flows based on their final destination and assume that congestion originates only at endpoints. Other solutions such as Regional Explicit Congestion Notification (RECN) [11] and Efficient and cost-effective Congestion-Control (EcoCC) [14] consider that congestion may originate also at internal points of the network. However, these techniques require additional resources such as mechanisms to detect and separate hot flows, specific control messages, and Content-Addressable Memories (CAM) to keep track of congested points at each port.…”

Section: Solutions To Hol Blockingmentioning

confidence: 99%

Combining Source-adaptive and Oblivious Routing with Congestion Control in High-performance Interconnects using Hybrid and Direct Topologies

Yébenes

Rocher-González

Escudero-Sahuquillo

et al. 2019

ACM Trans. Archit. Code Optim.

Self Cite

View full text Add to dashboard Cite

Hybrid and direct topologies are cost-efficient and scalable options to interconnect thousands of end nodes in high-performance computing (HPC) systems. They offer a rich path diversity, high bisection bandwidth, and a reduced diameter guaranteeing low latency. In these topologies, efficient deterministic routing algorithms can be used to balance smartly the traffic flows among the available routes. Unfortunately, congestion leads these networks to saturation, where the HoL blocking effect degrades their performance dramatically. Among the proposed solutions to deal with HoL blocking, the routing algorithms selecting alternative routes, such as adaptive and oblivious, can mitigate the congestion effects. Other techniques use queues to separate congested flows from non-congested ones, thus reducing the HoL blocking. In this article, we propose a new approach that reduces HoL blocking in hybrid and direct topologies using source-adaptive and oblivious routing. This approach also guarantees deadlock-freedom as it uses virtual networks to break potential cycles generated by the routing policy in the topology. Specifically, we propose two techniques, called Source-Adaptive Solution for Head-of-Line Blocking Avoidance (SASHA) and Oblivious Solution for Head-of-Line Blocking Avoidance (OSHA). Experiment results, carried out through simulations under different traffic scenarios, show that SASHA and OSHA can significantly reduce the HoL blocking.

show abstract

“…Como esses sistemas costumam usar Message Passing Interface (MPI) como padrão para desenvolvimento de aplicações paralelas, e as características de comunicação dessas aplicações variam conforme sua finalidade específica, a rede tem uma forte influência sobre o desempenho geral. Assim, a interconexão deve ser capaz de garantir uma alta largura de banda de comunicação, bem como baixa latência, para lidar com os requisitos das aplicações e não tornar-se o gargalo de todo o sistema [Escudero-Sahuquillo et al 2015, Kamburugamuve et al 2017.…”

Section: Introductionunclassified

O Impacto da Interconexão de Rede no Desempenho de Programas Paralelos

Maliszewski

Roloff

Griebler

et al. 2019

Anais Do XX Simpósio Em Sistemas Computacionais De Alto Desempenho (SSCAD 2019)

View full text Add to dashboard Cite

O desempenho de aplicações paralelas depende de dois componentes principais do ambiente; o poder de processamento e a interconexão de rede. Neste trabalho, foi avaliado o impacto de uma interconexão de alto desempenho em programas paralelos em um cluster homogêneo de servidores interconectados por Gigabit Ethernet 1 Gbps e InfiniBand FDR 56 Gbps. Foi realizada uma caracterização do NAS Parallel Benchmarks em relação à computação, comunicação e custo de execução em instâncias da Microsoft Azure. Os resultados mostraram que, em aplicações altamente dependentes de rede, o desempenho pode ser significativamente melhorado ao utilizar InfiniBand a um custo de execução melhor, mesmo com o preço superior da instância.

show abstract

Efficient and Cost-Effective Hybrid Congestion Control for HPC Interconnection Networks

Cited by 23 publications

References 15 publications

Application-aware Congestion Mitigation for High-Performance Computing Systems

Application-aware Congestion Mitigation for High-Performance Computing Systems

Combining Source-adaptive and Oblivious Routing with Congestion Control in High-performance Interconnects using Hybrid and Direct Topologies

O Impacto da Interconexão de Rede no Desempenho de Programas Paralelos

Contact Info

Product

Resources

About