The Surprising Impact of 1% Packet Loss

As part of this research, we demonstrate the surprising impact of 1% packet loss on throughput, both in symmetric and asymmetric networking topologies, in the environment using CUBIC congestion avoidance algorithm. Our findings reveal a significant decrease in throughput, more than 70%, compared to baseline measurements without packet loss. Moreover, we explore the effects of increasing packet loss levels, up to 10%, and observe a compounding decline in throughput, indicating the importance of addressing even minor levels of packet loss. We compare attained results using CUBIC congestion avoidance algorithm, in both topologies, with those achieved using the BBR congestion avoidance algorithm, advocating for broader and faster adoption of BBR.

Kemal Sanjta: Kemal is a result-oriented engineer focusing on designing, operating and troubleshooting large-scale networks. Passionate Linux user with a deep understanding of SRE/NRE practices. Over the last two decades, he worked at several large-scale companies applying NRE practices and automating remediation actions. As a Principal Internet Analyst at ThousandEyes, he focuses on research and providing deep and meaningful insights into outages through the lenses of ThousandEyes.

Speaker: Kemal Sanjta - Cisco ThousandEyes