Overview
Project Falco aims to decouple switching hardware and software to address latency issues in data center applications. The initiative led to the development of Pigeon, a 3.2Tbps switching platform, enhancing control over network features and performance.
What You'll Learn
1
How to identify and address microburst issues in network performance
2
Why decoupling hardware and software can improve network control
3
How to implement telemetry solutions for better network monitoring
Prerequisites & Requirements
- Understanding of network infrastructure and switching concepts
- Experience with network troubleshooting and performance analysis(optional)
Key Questions Answered
What is Project Falco and its main objectives?
Project Falco focuses on decoupling network hardware and software to address latency issues in data centers. By developing their own switching platform, Pigeon, LinkedIn aims to gain greater control over network features and improve performance without being dependent on commercial vendors.
What challenges did LinkedIn face with commercial switches?
LinkedIn encountered several challenges with commercial switches, including unaddressed software bugs, unnecessary features, lack of Linux-based automation tools, outdated monitoring software, and high costs associated with scaling licenses and support.
How does the Pigeon switch improve network performance?
The Pigeon switch, with a capacity of 3.2Tbps, addresses latency issues by allowing better control over telemetry data and programmability, enabling faster response to network demands and reducing the impact of microbursts on application performance.
What is the significance of telemetry in network management?
Telemetry is crucial for network management as it provides insights into performance metrics, such as buffer statistics, which help in predicting and mitigating issues like microbursts that can lead to latency problems in data centers.
Key Statistics & Figures
Switching capacity
3.2Tbps
This is the capacity of the Pigeon switch, which is designed to handle high-speed data transfers in LinkedIn's data centers.
Development hours invested
11,520 hours
This is the total time the Falco working group dedicated to developing the Pigeon switch.
Technologies & Tools
Some links below are affiliate links. We may earn a commission if you make a purchase.
Switching Platform
Pigeon
Developed to decouple switching hardware and software for better control and performance.
Messaging System
Kafka
Used for metrics and telemetry within the application layer of the switch.
Merchant Silicon
Tomahawk
The architecture of the Pigeon switch is based on the latest Tomahawk silicon.
Key Actionable Insights
1Implementing a decoupled network architecture can significantly enhance control over network performance and features.By developing solutions like Pigeon, organizations can tailor their network infrastructure to meet specific application demands, improving latency and overall efficiency.
2Utilizing telemetry data from switches can provide critical insights into network performance and help identify issues before they impact applications.Access to detailed telemetry allows for proactive management of network resources, which is essential in high-demand environments like data centers.
3Engaging with multiple merchant silicon vendors can lead to better customization and performance of network switches.This approach allows organizations to select the best silicon for their needs, ensuring that they can optimize their network infrastructure effectively.
Common Pitfalls
1
Relying solely on commercial vendors for network solutions can lead to delays in addressing critical issues.
This often results in prolonged latency problems, as seen with the microburst issue, highlighting the need for in-house solutions.
Related Concepts
Network Infrastructure Design
Latency Reduction Techniques
Programmable Data Centers
Telemetry In Network Management