On the impact of packet spraying in data center networks

329 citations recorded by the source corpus.

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

Presto: Edge-based Load Balancing for Fast Datacenter Networks (2015, background): “At network speeds of 10+ Gbps, current GRO algorithms are unable to sustain line rate under severe reordering due to extreme computational overhead, and hence per-packet load balancing approaches [17, 22] need to be reconsidered.”
pFabric: minimal near-optimal datacenter transport (2013, methodology): “Fabric load-balancing: We use packet spraying [11], where each switch sprays packets among all shortest-path next hops in roundrobin fashion.”
Homa: a receiver-driven low-latency transport protocol using network priorities (2018, methodology): “ latency is to eliminate queuing delays. Similar to prior work, we assume that bandwidth in the network core is sufﬁcienttoaccommodatetheofferedload,andthatthenetwork supports efﬁcient load-balancing [10,16,1], so that packets are distributed evenly across the available paths (we assume simple rando”
Re-architecting datacenter networks and stacks for low latency and high performance (2017, background): “TCP) and in-network packet-spraying results in heavily reduced performance in these scenarios, because the transport protocol is not aware that only one of its paths is misbehaving [12, 31].”
DRILL: Micro Load Balancing for Low-latency Data Center Networks (2017, background): “category include balancing “flowlets” [19, 44] or per-packet spreading of traffic in a round robin fashion [27, 34].”