A scalable, commodity data center network architecture
@inproceedings{AlFares2008ASC, title={A scalable, commodity data center network architecture}, author={Mohammad Al-Fares and Alexander Loukissas and Amin Vahdat}, booktitle={SIGCOMM '08}, year={2008} }
Today's data centers may contain tens of thousands of computers with significant aggregate bandwidth requirements. The network architecture typically consists of a tree of routing and switching elements with progressively more specialized and expensive equipment moving up the network hierarchy. Unfortunately, even when deploying the highest-end IP switches/routers, resulting topologies may only support 50% of the aggregate bandwidth available at the edge of the network, while still incurring…
Figures and Tables from this paper
3,251 Citations
MCube: A high performance and fault-tolerant network architecture for data centers
- Computer Science2010 International Conference On Computer Design and Applications
- 2010
MCube is a server-centric network architecture specifically designed for data centers that leverages commodity mini-switches to provide high aggregate bandwidth and can deliver more performance at less cost than available from today's higher-end solutions.
Data Center Switch Architecture in the Age of Merchant Silicon
- Computer Science2009 17th IEEE Symposium on High Performance Interconnects
- 2009
This paper describes how to save cost and power by repackaging an entire data center network as a distributed multi-stage switch using a fat-tree topology and merchant silicon instead of proprietary ASICs.
Scalable and Cost-Effective Interconnection of Data-Center Servers Using Dual Server Ports
- Computer ScienceIEEE/ACM Transactions on Networking
- 2011
It is proven that FiConn is highly scalable to encompass hundreds of thousands of servers with low diameter and high bisection width, and developed a low-overhead traffic-aware routing mechanism to improve effective link utilization based on dynamic traffic state.
SPAIN : Design and Algorithms for Constructing Large DataCenter Ethernets from Commodity Switches
- Computer Science
- 2009
SPAIN provides multipath forwarding using inexpensive, commodity off-the-shelf (COTS) switches, over arbitrary topologies, and demonstrates SPAIN’s ability to improve bisection bandwidth over both simulated and real data-center networks.
Impact of network protocols on data center applications
- Computer Science
- 2013
Evaluating the advantages of jumbo frames using modern hardware with features such as large send/receive offload, and with canonical data center applications such as MapReduce and tiered Web services, finds that the throughput and CPU utilization benefits still exist generally, although compared to prior studies, are significantly reduced.
LEGUP: using heterogeneity to reduce the cost of data center network upgrades
- Computer ScienceCoNEXT
- 2010
The theory of heterogeneous Clos networks is developed and it is shown that the construction needs only as much link capacity as the classic Clos network to route the same traffic matrices and this bound is the optimal.
SPAIN: COTS Data-Center Ethernet for Multipathing over Arbitrary Topologies
- Computer ScienceNSDI
- 2010
SPAIN ("Smart Path Assignment In Networks") provides multipath forwarding using inexpensive, commodity off-the-shelf (COTS) Ethernet switches, over arbitrary topologies, and is demonstrated to improve bisection bandwidth over both simulated and experimental data-center networks.
Dahu: Commodity switches for direct connect data center networks
- Computer ScienceArchitectures for Networking and Communications Systems
- 2013
This work presents Dahu1, simple enhancements to commodity Ethernet switches to support direct connect networks in data centers, which avoids congestion hot-spots by dynamically spreading traffic uniformly across links, and forwarding traffic over non-minimal paths where possible.
PAST: scalable ethernet for data centers
- Computer ScienceCoNEXT '12
- 2012
Surprisingly, PAST can achieve performance comparable to or greater than Equal-Cost Multipath (ECMP) forwarding, which is currently limited to layer-3 IP networks, without any multipath hardware support.
A High Performance, Low-Cost Network Architecture for Data Centers
- Computer Science2010 International Conference on Cyber-Enabled Distributed Computing and Knowledge Discovery
- 2010
A new network interconnection structure for data centers is presented, which leverages largely low-cost commodity Ethernet switches to support the full aggregate bandwidth of the network.
References
SHOWING 1-10 OF 42 REFERENCES
Dynamo: amazon's highly available key-value store
- Computer ScienceSOSP
- 2007
D Dynamo is presented, a highly available key-value storage system that some of Amazon's core services use to provide an "always-on" experience and makes extensive use of object versioning and application-assisted conflict resolution in a manner that provides a novel interface for developers to use.
Fat-trees: Universal networks for hardware-efficient supercomputing
- Computer ScienceIEEE Transactions on Computers
- 1985
The author presents a new class of universal routing networks, called fat-trees, which might be used to interconnect the processors of a general-purpose parallel supercomputer, and proves that a fat-tree of a given size is nearly the best routing network of that size.
The Click modular router
- Computer ScienceSOSP
- 1999
The Click IP router can forward 64-byte packets at 73,000 packets per second, just 10% slower than Linux alone, and is easy to extend by adding additional elements, which are demonstrated with augmented configurations.
Tree bitmap: hardware/software IP lookups with incremental updates
- Computer ScienceCCRV
- 2004
This paper presents a family of IP lookup schemes using a data structure that compactly encodes large prefix tables in order to address the criteria listed above and presents a series of optimizations to the core algorithm that allows the memory access width of the algorithm to be reduced at the cost of memory references or allocated memory.
Myrinet: A Gigabit-per-Second Local Area Network
- Computer ScienceIEEE Micro
- 1995
The Myrinet local area network employs the same technology used for packet communication and switching within massively parallel processors, but with the highest performance per unit cost of any current LAN.
Faster IP lookups using controlled prefix expansion
- Computer ScienceSIGMETRICS '98/PERFORMANCE '98
- 1998
It is described how IP lookups can be made faster using a new technique called controlled prefix expansion, together with optimization techniques based on dynamic programming, to improve the speed of the best known IP lookup algorithms by at least a factor of two.
Fast address lookups using controlled prefix expansion
- Computer ScienceTOCS
- 1999
The main technique, controlled prefix expansion, transforms a set of prefixes into an equivalent set with fewer prefix lengths, and optimization techniques based on dynamic programming, and local transformations of data structures to improve cache behavior are used.
IBM Research Report Design and Analysis of the BlueGene/L Torus Interconnection Network
- Computer Science
- 2003
A parallel performance simulator that was used extensively to help architect and design the torus network is described and sample simulator performance studies that contributed to design decisions are presented.
Statistical bandwidth sharing: a study of congestion at flow level
- BusinessSIGCOMM '01
- 2001
The statistics of the realized throughput of elastic document transfers are studied, accounting for the way network bandwidth is shared dynamically between the randomly varying number of concurrent flows.
The Network Architecture of the Connection Machine CM-5
- Computer Science, BusinessJ. Parallel Distributed Comput.
- 1996
The Connection Machine Model CM-5 Supercomputer is a massively parallel computer system designed to offer performance in the range of 1 teraflops (1012floating-point operations per second). The CM-5…