Improving datacenter performance and robustness with multipath TCP

Raiciu, Costin; BarreSebastien,; PluntkeChristopher,; GreenhalghAdam,; WischikDamon,; HandleyMark,

doi:10.1145/2043164.2018467

Cited by 262 publications

(116 citation statements)

References 6 publications

Supporting

Mentioning

108

Contrasting

Order By: Relevance

“…MPTCP is likely to work correctly in the Internet through different middleboxes [28]. Study in [29] shows robust performance of MPTCP with a few subflows per-flow in a multiple fat-tree-topology data center environment. But the related buffer sizing issues are not discussed.…”

Section: Background and Related Workmentioning

confidence: 99%

Exploring parallelism and desynchronization of TCP over high speed networks with tiny buffers

Xue

Chiu

Kondikoppa

et al. 2015

Computer Communications

View full text Add to dashboard Cite

a b s t r a c tThe buffer sizing problem is a big challenge for high speed network routers to reduce buffer cost without throughput loss. The past few years have witnessed debate on how to improve link utilization of high speed networks where the router buffer size is idealized into dozens of packets. Theoretically, the buffer size can be shrunk by more than 100 times. Under this scenario, widely argued proposals for TCP traffic to achieve acceptable link capacities mandate three necessary conditions: over-provisioned core link bandwidth, nonbursty flows, and tens of thousands of asynchronous flows. However, in high speed networks where these conditions are insufficient, TCP traffic suffers severely from routers with tiny buffers.To explore better performance, we propose a new congestion control algorithm called Desynchronized Multi-Channel TCP (DMCTCP) that creates a flow with parallel channels. These channels can desynchronize each other to avoid TCP loss synchronization, and they can avoid traffic penalties from burst losses. Over a 10 Gb/s large delay network ruled by tiny buffer routers, our emulation results show that bottleneck link utilization can reach over 80% with much fewer number of flows. Compared with other TCP congestion control variants, DMCTCP can also achieve much better performance in high loss rate networks. Facing the buffer sizing challenge, our study is a new step towards the deployment of optical packet switching networks.

show abstract

Section: Background and Related Workmentioning

confidence: 99%

Exploring parallelism and desynchronization of TCP over high speed networks with tiny buffers

Xue

Chiu

Kondikoppa

et al. 2015

Computer Communications

View full text Add to dashboard Cite

show abstract

“…To maximize the utility of such an advantage, a switch path can be utilized as k weak parallel paths, which share the same set of switches but have different intermediate servers. Such parallel and weak parallel paths between any pair of servers can be further utilized to improve the transmission rate or to enhance the transmission reliability for one-to-one traffic with Multipath TCP [26,27]. In addition, Multipath TCP can explore such multiple paths to tackle traffic congestion, leading to higher network utilization.…”

Section: Multi-paths For One-to-one Trafficmentioning

confidence: 99%

DCube: A family of network structures for containerized data centers using dual-port servers

Guo

Li²,

et al. 2014

Computer Communications

View full text Add to dashboard Cite

a b s t r a c tA fundamental goal of datacenter networking is to efficiently interconnect a large number of servers in a cost-effective way. Inspired by the commodity servers in today's data centers that come with dual-port, we consider how to design low-cost, robust, and symmetrical network structures for containerized data centers with dual-port servers and low-end switches. In this paper, we propose a family of such network structure called a DCube, including H-DCube and M-DCube. The DCube consists of one or multiple interconnected sub-networks, each of which is a compound graph made by interconnecting a certain number of basic building blocks by means of a hypercube-like graph. More precisely, the H-DCube and M-DCube utilize the hypercube and 1-möbius cube, respectively, while the M-DCube achieves a considerably higher aggregate bottleneck throughput compared to H-DCube. Mathematical analysis and simulation results show that the DCube exhibits graceful performance degradation as the failure rate of server or switch increases. Moreover, the DCube significantly reduces the required wires and switches compared to the BCube and fat-tree. In addition, the DCube achieves a higher speedup than the BCube does for the one-to-several traffic patterns. The proposed methodologies in this paper can be applied to the compound graph of the basic building block and other hypercube-like graphs, such as Twisted cube, Flip MCube, and fastcube.

show abstract

“…That is, they can install custom multipath routing protocols, like Multipath TCP [37] and SPAIN [32], on all end-hosts. This assumption does not hold in cloud settings, where customers can run their own OS installations.…”

Section: B Switches Links and End-hostsmentioning

confidence: 99%

“…This assumption is not valid when using single-path protocols like spanning tree; however, near-optimal loadbalancing can be achieved on arbitrary topologies by using Multipath TCP [37] or SPAIN [32]. Based on their evaluations, we believe that these multipath protocols would be able to fully utilize REWIRE's constructions.…”

Section: Operating An Unstructured Dcnmentioning

confidence: 99%

REWIRE: An optimization-based framework for unstructured data center network design

Curtis

Carpenter

Elsheikh

et al. 2012

2012 Proceedings IEEE INFOCOM

View full text Add to dashboard Cite

Despite the many proposals for data center network (DCN) architectures, designing a DCN remains challenging. DCN design is especially difficult when expanding an existing network, because traditional DCN design places strict constraints on the topology (e.g., a fat-tree). Recent advances in routing protocols allow data center servers to fully utilize arbitrary networks, so there is no need to require restricted, regular topologies in the data center. Therefore, we propose a data center network design framework, that we call REWIRE, to design networks using an optimization algorithm. Our algorithm finds a network with maximal bisection bandwidth and minimal end-to-end latency while meeting user-defined constraints and accurately modeling the predicted cost of the network. We evaluate REWIRE on a wide range of inputs and find that it significantly outperforms previous solutions-its network designs have up to 100-500% more bisection bandwidth and less end-to-end network latency than equivalent-cost DCNs built with best practices.

show abstract

Improving datacenter performance and robustness with multipath TCP

Cited by 262 publications

References 6 publications

Exploring parallelism and desynchronization of TCP over high speed networks with tiny buffers

Exploring parallelism and desynchronization of TCP over high speed networks with tiny buffers

DCube: A family of network structures for containerized data centers using dual-port servers

REWIRE: An optimization-based framework for unstructured data center network design

Contact Info

Product

Resources

About