IncBricks: Toward In-Network Computation with an In-Network Cache

176 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.

PyTorch FSDP: Experiences on Scaling Fully Sharded Data Parallel (2023, background): “Specifically, datacenters typically adopt a fat-tree network topology [16] with over-subscription, leading to abundant locality to exploit and a well-motivated reason to reduce cross-host traffic [17].”
ATP: In-network Aggregation for Multi-tenant Learning (2021, background): “These limits lead to small packet sizes for in-network computation and storage applications: the payload size of SwitchML and NetCache is 128B [40, 41, 46, 56] 1.”
NetCache: Balancing Key-Value Stores with Fast In-Network Caching (2017, background): “IncBricks [31] designs middleboxes to cache items inside the network.”
Accelerometer: Understanding Acceleration Opportunities for Data Center Overheads at Hyperscale (2020, background): “Examples include remote ML inference units [51], network switches [78, 104], remote encryption units [31], and remote GPUs [43].”
NetChain: Scale-Free Sub-RTT Coordination (2018, unknown): “IncBricks [49] caches key-value items with NPUs.”
Scaling Distributed Machine Learning with In-Network Aggregation (2019, unknown): “als [20,21,43,44,50,56] to offload, when appropriate [73], application-specific primitives into network devices.”