Ethernet as a Lossless Deadlock Free System Area Network

Data centers must support a range of workloads with differing demands. Although existing approaches handle routine traffic smoothly, intense hotspots-even if ephemeralcause excessive packet loss and severely degrade performance. This loss occurs even though congestion is typically highly localized, with spare buffer capacity at nearby switches. In this paper, we argue that switches should share buffer capacity to effectively handle this spot congestion without the monetary hit of deploying large buffers at individual switches. Specifically, we present detour-induced buffer sharing (DIBS), a mechanism that achieves a near lossless network without requiring additional buffers at individual switches. Using DIBS, a congested switch detours packets randomly to neighboring switches to avoid dropping the packets. We implement DIBS in hardware, on software routers in a testbed, and in simulation, and we demonstrate that it reduces the 99th percentile of delay-sensitive query completion time by up to 85%, with very little impact on other traffic.

show abstract

“…DIBS is also free of problems such as deadlock [22], as we do not require any host or switch to stop transmitting.…”

Section: Related Workmentioning

confidence: 97%

Dibs

Zarifis

Miao

Calder

et al. 2014

Proceedings of the Ninth European Conference on Computer Systems

View full text Add to dashboard Cite

show abstract

“…technologies, including Ethernet [8]. It is a spanning tree based routing algorithm that allows the use of all links only constrained by turn prohibition [18], opposed to to the link prohibition used in STP.…”

Section: Related Workmentioning

confidence: 99%

“…Flow control is mandatory to avoid frame loss, and especially important in a Storage Area Network where protocols such as iSCSI [4] and ATA over Ethernet [5] are deployed. Flow control also improves the performance of higher layer protocols such as TCP [6], [7], [8]. Virtual channels have many purposes and can be used for traffic segregation, QoS, and deadlock free shortest path routing [9].…”

Section: Introductionmentioning

confidence: 99%

Effective Shortest Path Routing for Gigabit Ethernet

Reinemo

Skeie

2007

2007 IEEE International Conference on Communications

View full text Add to dashboard Cite

Ethernet's native routing algorithm, the Spanning Tree Protocol, causes a major performance bottleneck when network connectivity increases. Since the Spanning Tree Protocol avoids deadlocks and infinitely looping packets by turning any topology into a tree, it leaves a large portion of links unused and thus wastes bandwidth. In this paper we address this weakness by proposing a new routing algorithm that avoids disabling links and prohibiting turns, and that guarantees shortest path routing.Through the use of layered routing we show how to improve performance with respect to both the Spanning Tree Protocol and a more recent proposal called Tree-Based Turn-Prohibition. The concept of layered routing is used to group virtual channels into network layers, and assign a limited set of source/destination address pairs to each layer. Extensive simulations show that we are able to increase throughput by a factor of more than 3.5 compared to the Spanning Tree Protocol and a factor of 1.8 compared to Tree-Based Turn-Prohibition. Our concept relies on features introduced in IEEE standards 802.1Q, 802.1D and 802.3x, as well as changes currently discussed in IEEE task forces.

show abstract

“…Another deterministic algorithm suitable for flowcontrolled Ethernet technology is the Up*/Down* algorithm (UD) [23]. UD can be used in any topology without the need for virtual channels (not supported by Ethernet), making it suitable for a wide range of network technologies, including Ethernet [21]. UD first selects a root node, then it computes a spanning tree, and finally it enforces turn prohibitions among the links to avoid cycles.…”

Section: Routing Algorithms For Ethernetmentioning

confidence: 99%

“…Zero copy protocols [3,27] and remote memory access techniques [19] have been added to Ethernet interfaces in order to improve latency. Also, the addition of an on/off based flow-control mechanism to the Ethernet standard in 1997 [24], allows Ethernet to behave as a loss-less network [21] without the need for expensive retransmission and acknowledgement of packets. The only major problem left is the routing algorithm, which only supports the tree topology.…”

Section: Introductionmentioning

confidence: 99%