Beyond fat-trees without antennae, mirrors, and disco-balls

Kassing, Simon; Valadarsky, Asaf; Shahaf, Gal; Schapira, Michael; Singla, Ankit

doi:10.1145/3098822.3098836

Cited by 80 publications

(96 citation statements)

References 24 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…The introduction of flexibility enables a formal argument for and against flexibility of systems. This shall be demonstrated on the argumentative line of [6]. In this work, the authors argue in favor of static network topologies in data centers, which they claim to be in fact more flexible than reconfigurable topologies that use optical or radio links to bypass a fat-tree topology.…”

Section: Traffic Flexibility In Data Centersmentioning

confidence: 90%

“…The works that directly discuss or target flexibility itself, and that we are aware of, can be reduced to [5]- [8]. In [5], [6] the authors discuss the flexibility of traffic engineering solutions in data centers, arguing against the assumption that wired connections are inflexible. To support their argument, the used flexibility metric is the throughput performance for increasing traffic, compared to a "throughput proportional" behavior.…”

Section: Related Workmentioning

confidence: 99%

“…Hence, with such a sequenceĈ a static implementation can realize all demand changes that an adaptive can. Hence, it is considered to be more flexible in this respect, as stated in [6] as well. However, considering the whole spectrum of traffic matrices, investigating different setups or introducing strict timing or low budget constraints might give different order on the flexibility of these approaches.…”

Section: Traffic Flexibility In Data Centersmentioning

confidence: 99%

See 2 more Smart Citations

A Mathematical Measure for Flexibility in Communication Networks

Klügel

Kellerer

et al. 2019

2019 IFIP Networking Conference (IFIP Networking)

View full text Add to dashboard Cite

For communication networks research, flexibility of network design and networking solutions is considered a competitive advantage. However, this advantage is typically only claimed on an argumentative level and neither formally supported nor thoroughly investigated. To support the claim of flexibility and to make the flexibility of different solutions comparable, its degree must be quantified and thus made measurable. In this work, we propose a mathematical basis to quantify a degree of flexibility achieved by communication networks. We motivate that flexibility can be cast to the "size" of a set of achievable demand changes. Consequently, we propose the use of mathematical measure theory to quantify achieved networking flexibility. We derive several implications on the basic structure of flexibility, extend the insights towards a utility of flexibility and develop systematic approaches for both analytical and empirical measurement of flexibility. We apply the insights to several use-cases, showing that flexibility is in fact not as straight-forward to argue as it seems at first glance.

show abstract

Section: Traffic Flexibility In Data Centersmentioning

confidence: 90%

Section: Related Workmentioning

confidence: 99%

Section: Traffic Flexibility In Data Centersmentioning

confidence: 99%

See 1 more Smart Citation

A Mathematical Measure for Flexibility in Communication Networks

Klügel

Kellerer

et al. 2019

2019 IFIP Networking Conference (IFIP Networking)

View full text Add to dashboard Cite

show abstract

“…However, as a high-level term, its interpretation was different across various research fields and use cases. For example, in DC traffic engineering, the authors in [24] consider it as the additional throughput performance for increasing traffic, in comparison to a proportional behavior. In [8], we propose a general measure of flexibility that is abstracted from different use cases.…”

Section: Toward Designing Flexible Control Planementioning

confidence: 99%

Optimizing the Flexibility of SDN Control Plane

Huang

Kellerer

2020

NOMS 2020 - 2020 IEEE/IFIP Network Operations and Management Symposium

View full text Add to dashboard Cite

Facing fast-changing network traffic, Software Defined Networking (SDN) endows the capability of efficient traffic forwarding and control plane resource management, which results in flexibility in comparison to networking with legacy rigid hardware. Meanwhile, flexibility has become an implicit target of many novel network algorithms and designs. This paper answers an interesting research question: "Can we optimize the flexibility as an objective?" We study the impact of Data Centers (DCs) location on the flexibility of dynamic control plane. The flexibility is revealed when the control plane can adapt itself with controller migration and switch re-assignment in a timely manner for a new group of flows to satisfy the requirements of flow setup. We propose a model, named FLEXDC, for static DC placement and dynamic controller placement to optimize the control plane's flexibility. We also design heuristics to speed up the decision process. Our simulation over real network topology with synthetic flows shows the improved flexibility of the dynamic control plane. Furthermore, we can save up to 2 DCs while achieving the same flexibility, compared with a naive approach.

show abstract

“…As leaf-spine networks have shorter paths than 3-tier Clos networks, it is not clear if the gains of expanders apply to leaf-spine networks. In addition, current expander designs require uncommon or novel transport, routing, or forwarding protocols, like MPTCP with -shortest path routing [23], or an ECMP/VLB routing hybrid with dynamic switching at flowlet granularity [15]. Using such protocols is always a deployment hurdle, but would be a non-starter for enterprises that depend on standard hardware and IT support, lacking the hyperscale operators' ability to develop and support custom designs.…”

mentioning

confidence: 99%

Spineless Data Centers

Harsh

Jyothi

Godfrey

2020

Proceedings of the 19th ACM Workshop on Hot Topics in Networks

View full text Add to dashboard Cite

In enterprises, CDNs, and increasingly in edge computing, most data centers have moderate scale. Recent research has developed designs such as expander graphs that are highly efficient compared to largescale, 3-tier Clos networks, but moderate-scale data centers need to be constructed with standard hardware and protocols familiar to network engineers, and are overwhelmingly built with a leaf-spine architecture. This paper explores whether the performance efficiency that is known to be theoretically possible at large scale can be realized in a practical way for the common leaf-spine data center. First, we find that more efficient topologies indeed exist at moderate scale, showing through simulation and analysis that much of the benefit comes from choosing a "flat" network that uses one type of switch rather than having separate roles for leafs and spines; indeed, even a simple ring-based topology outperforms leaf-spine for a wide range of traffic scenarios. Second, we design and prototype an efficient routing scheme for flat networks that uses entirely standard hardware and protocols. Our work opens new research directions in topology and routing design that can have significant impact for the most common data centers. CCS CONCEPTS • Networks → Network design principles; Routing protocols.

show abstract

Beyond fat-trees without antennae, mirrors, and disco-balls

Cited by 80 publications

References 24 publications

A Mathematical Measure for Flexibility in Communication Networks

A Mathematical Measure for Flexibility in Communication Networks

Optimizing the Flexibility of SDN Control Plane

Spineless Data Centers

Contact Info

Product

Resources

About