Yiran Blog

New post every day (with probability 0.03).

In-band Network Telemetry (INT)

In-band Network Telemetry

In-band Network Telemetry(Specification) Towards Knowledge-Defined Networking using In-band Network Telemetry INTCollector: A High-performance Collector for In-band Network Telemetry

Posted by Yiran on March 31, 2019

2D CoNEXT 2018

Workload Adaptive Flow Scheduling

核心思想 实现不同特征workload下的adaptive scheduling, 而以前的调度策略只针对特定特征的workload. 2D指的是 multiplexing (比如fair sharing) 和 serialization (比如FIFO SRPT). Observations and Insights Within-workload variabil...

Posted by Yiran on March 31, 2019

Dapper SOSR 2017

Data plane performance diagnosis of TCP

核心思想 针对多租户场景在 network edge (e.g., hypervisor, NIC, or ToR switch) monitor TCP性能，诊断出一个connection是server-limited，network-limited还是receiver-limited. 主要挑战 需要在硬件（SmartNIC）上减少per-flow state存储开销 Des...

Posted by Yiran on March 30, 2019

PASE SIGCOMM 2014

经典论文：对dc transport 总结

核心思想 对已有dc transport 工作总结，结合各类方法的优势提出新的传输优化。 已有工作分类 （1）Self adjusting endpoints：发送端独立地根据自己探测到的网络状况调整速率，TCP，DCTCP等都属于此类。 网络拥塞状况通过拥塞信号获知，包括隐式的（丢包）和显式的（ECN）。当拥塞发生时，窗口会被减少，如果以追求公平性为目标，所有流的窗口都减相同的fa...

Posted by Yiran on March 29, 2019

NetBouncer NSDI 2019

Failure Localization in Data Center Networks

核心思想 通过IP-in-IP 的probe探测，定位数据中心故障，包括device故障和link故障。（微软） Network Troubleshooting三个requirement(Motivation) End host’s perspective: 交换机无法观察到gray failure. gray failure: 交换机不会记录. e.g. drop packet...

Posted by Yiran on March 28, 2019

Cloud Monitoring IMC 2018

Experiences and Challenges

核心思想 指出VNET的monitor和physical网路的monitor不同 提出三个问题 数据中心运营者能monitor VNET性能吗？之前那些用于物理网络的工具（Pingmesh，Everflow等）能够适用于VNET吗？ 已有的monitor方法在测量VNET性能时准确度多高？能够detect出customer-impacting的问题吗？precision和rec...

Posted by Yiran on March 28, 2019

深度增强学习在数据中心网络中应用

论文汇总 Keep Updating

深度增强学习在数据中心网络中应用 Keep Updating Classic Meets Modern: a Pragmatic Learning-Based Congestion Control for the Internet (SIGCOMM 2020) Server-Driven Video Streaming for Deep Learning Inference (SIGCO...

Posted by Yiran on March 27, 2019

Useful links

(Keep Updating) Research and Paper Writing, Miscellaneous

Research and Paper Writing How to Look for Ideas in Computer Science Research How to Do Great Research (Grad school survival advice from Nick Feamster and Alex Gray) How to Read Writing Te...

Posted by Yiran on March 27, 2019

Timely SIGCOMM 2015

RTT-based Congestion Control

核心思想 第一个RTT-based congestion control在数据中心的应用，认为RTT是个很好的拥塞信号。 拥塞信号的信息量 传统TCP的ECN ————> 超过一定阈值才标记，只有0 和 1的变化 DCTCP的ECN ————> 更进了一步，将单bit转化为一个RTT内的多bit 但是ECN的方式对多优先级不general：低优先级不被标...

Posted by Yiran on March 27, 2019

Thoughts of Programmable Switches SIGCOMM 2018 CCR

The Role of Programmable Switches

核心思想 在可编程交换机上实现的有两类：应用功能（Load Balancer比如SilkRoad、Load partition比如NetCache） 和 网络功能 (Network traffic balancing、Network Telemetry、Packet scheduling、Congestion Control) 其实几乎所有的应用的offload都可以通过serv...

Posted by Yiran on March 27, 2019

FEATURED TAGS
Measurement in Datacenter TCP Performance Diagnosis Flow Scheduling Machine Learning Network Troubleshooting Programmable Switches Congestion Control Useful links Deep Reinforcement Learning Transport in Datacenter Key-value Store Load Balancing Distributed Systems OS Scheduling microsecond-scale
ABOUT ME

✉️ yiran914@gmail.com