Fat-trees: Universal networks for hardware-efficient supercomputing

Leiserson, Charles E.

doi:10.1109/tc.1985.6312192

Cited by 1,054 publications

(447 citation statements)

References 15 publications

Supporting

Mentioning

438

Contrasting

Unclassified

Order By: Relevance

“…5 Topology 3 is a simple linear topology of N =100 switches, with a host attached to each switch. Finally, Topology 4 is a small fat tree [33], as often used in data centre networks. The topology has 20 switches and a total of 80 switch ports, and is shown in Figure 5.…”

Section: Methodsmentioning

confidence: 99%

Efficient topology discovery in OpenFlow-based Software Defined Networks

Pakzad

Portmann

Tan

et al. 2016

Computer Communications

View full text Add to dashboard Cite

Software Defined Networking (SDN) is a new networking paradigm, with a great potential to increase network efficiency, ease the complexity of network control and management, and accelerate the rate of technology innovation. One of the core concepts of SDN is the separation of the network's control and data plane. The intelligence and the control of the network operation and management, such as routing, is removed from the forwarding elements (switches) and is concentrated in a logically centralised component, i.e. the SDN controller. In order for the controller to configure and manage the network, it needs to have up-to-date information about the state of the network, in particular its topology. Consequently, topology discovery is a critical component of any Software Defined Network architecture. In this paper, we evaluate the efficiency of the de facto standard approach to topology discovery currently implemented by the major SDN controller frameworks, and propose simple and practical modifications, which achieve a significantly improved efficiency and reduced control overhead. We have implemented our new topology discovery approach on the widely used POX controller platform, and have evaluated it for a range of network topologies via experiments using the Mininet network emulator as well as a specific topology in the OFELIA SDN testbed. Our results show that our proposed modifications achieve an up to 40% reduction in controller load compared to the current state-of-the-art approach, while delivering identical discovery functionality.

show abstract

Section: Methodsmentioning

confidence: 99%

Efficient topology discovery in OpenFlow-based Software Defined Networks

Pakzad

Portmann

Tan

et al. 2016

Computer Communications

View full text Add to dashboard Cite

show abstract

“…Multi-pathing: Big Data and High Performance Computing applications have led to the creation of low oversubscription, high-capacity network interconnects like Clos [5], Fat-tree [6] etc. The data center network should be able to leverage these high-capacity interconnects by spreading traffic across the redundant links in these topologies and achieve high bisection bandwidth.…”

Section: Virtual Machine Migrationmentioning

confidence: 99%

ProActive routing in scalable data centers with PARIS

Arora

Benson

Rexford

2014

Proceedings of the 2014 ACM SIGCOMM Workshop on Distributed Cloud Computing

View full text Add to dashboard Cite

Modern data centers must meet many challenges and expectations: (a) They must scale to a large number of servers, while offering high bisection bandwidth and flexible placement of virtual machines.(b) They must allow tenants to specify network policies and realize them by forwarding traffic through the desired sequences of middleboxes. (c) They must allow tenants to bring their own IP address space to the cloud to ease transition of enterprise applications and middlebox configuration.The traditional approach of connecting layer-two pods through a layer-three core constrains VM placement. More recent "flat" designs are more flexible but have scalability limitations due to flooding/broadcasting or querying directories of VM locations. Rather than reactively learn VM locations, our PARIS architecture has a controller that pre-positions IP forwarding entries in the switches. Switches within a pod have complete information about the VMs beneath them, while each core switch maintains complete forwarding state for part of the address space. PARIS offers network designers the flexibility to choose a topology that meets their latency and bandwidth requirements.PARIS also allows tenants to bring their own IP address space and utilizes lightweight virtualization using Linux namespaces to offer middlebox service in a manner that truly reflects the pay-as-you go model of cloud computing. It utilizes MPLS label forwarding and aggregation in the network core and uses source routing at the network edge to realize middlebox policies specified by tenants. Finally, we evaluate our PARIS prototype built using Openflow-compliant switches and NOX controller.Using PARIS we can build a data center network that can support up to 500K VMs.iii

show abstract

“…In the rest of this paper, we shall apply the many sources regime results to analyzing buffer depth in a popular high radix topology -Clos [2] or fat tree networks [7]. The high radix switch queues are an appropriate application for the many sources asymptotic results.…”

Section: Application To High Radix Fat Treesmentioning

confidence: 99%

Buffer and Delay Bounds in High Radix Interconnection Networks

Singh

Dally²

2004

IEEE Comput. Arch. Lett.

View full text Add to dashboard Cite

Abstract-We apply recent results in queueing theory to propose a methodology for bounding the buffer depth and packet delay in high radix interconnection networks. While most work in interconnection networks has been focused on the throughput and average latency in such systems, few studies have been done providing statistical guarantees for buffer depth and packet delays. These parameters are key in the design and performance of a network. We present a methodology for calculating such bounds for a practical high radix network and through extensive simulations show its effectiveness for both bursty and non-bursty injection traffic. Our results suggest that modest speedups and buffer depths enable reliable networks without flow control to be constructed.

show abstract

Fat-trees: Universal networks for hardware-efficient supercomputing

Cited by 1,054 publications

References 15 publications

Efficient topology discovery in OpenFlow-based Software Defined Networks

Efficient topology discovery in OpenFlow-based Software Defined Networks

ProActive routing in scalable data centers with PARIS

Buffer and Delay Bounds in High Radix Interconnection Networks

Contact Info

Product

Resources

About