Pingmesh: A Large-Scale System for Data Center Network Latency Measurement and Analysis

By Chuanxiong Guo, Lihua Yuan, Dong Xiang, Yingnong Dang, Ray Huang, David A. Maltz, Zhaoyi Liu, Vin Wang, Bin Pang, Huaimo Chen and 2 more. Conference on Applications, Technologies, Architectures, and Protocols for Computer Communication, 2015

542 citations recorded by the source corpus.

Subfields: Computer Networks and Communications, Information Systems

View Paper: https://www.semanticscholar.org/paper/65da29a03c8905cbc0614612d1632864336c4786

How researchers see this paper

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

In Computer Networks and Communications

In Information Systems

Citations in the corpus

RDMA over Commodity Ethernet at Scale (2016, methodology): “We have developed an active latency measurement service for RDMA similar to the TCP Pingmesh service [21].”
MegaScale: Scaling Large Language Model Training to More Than 10, 000 GPUs (2024, background): “We believe that our work not only offers practical insights for those working on LLM training, but also paves the way for future research in this rapidly evolving field.”
Serverless Computing: One Step Forward, Two Steps Back (2018, background): “…close to typical intra-rack datacenter network measurements; studies from even a few years ago report average inter-rack measurements around 1.26ms [8] In sum, communicating via cloud storage is not a reasonable replacement for directly-addressed networking, even with direct I/O—it is at leaste…”
Datacenter RPCs can be General and Fast (2018, background): “This is not a major de ciency because datacenter networks typically use ECMP for load balancing, which preserves intra- ow ordering [67, 29, 66].”
FlowRadar: A Better NetFlow for Data Centers (2016, background): “There are also many end-host based monitoring solutions such as SNAP which captures TCP-level statistics [41] and pingmesh [25] which leverages active probes.”
Language-Directed Hardware Design for Network Performance Monitoring (2017, background): “For example, it is challenging to localize queue buildup to a particular switch or pinpoint traffic causing the queue buildup, forcing operators to infer the network-level root causes indirectly [40].”