Technical Tip: How to Test 10G Link Performance on a FortiGate

[wcruvinel]

Description	When testing the performance of a 10Gbps link, it is important to understand the complexity involved. This article offers insights into the complexities of testing 10G links and provides recommendations for accurate performance assessments. It is recommended to read 'Technical Tip: Low Throughput Troubleshooting' first, as it is an essential first step for delving deeper into performance testing of links.
Scope	FortiGate.
Solution	When testing a 10G link, keep the following in mind: Testing Tools vs. Everyday Use: There are special tools, such as iperf, that can test the maximum speed of a link. But in daily activities such as downloading files, the full 10G speed may not always be used. What Affects Speed: Many things can change the speed obtained. This includes how fast the server being downloaded from can send data, and the 'back-and-forth' nature of how data is sent over the internet. How Data is Sent Over the Internet: When downloading a file, a lot of checking is performed. The computer sending the data waits for a 'thumbs up' from the computer before sending more. This process can slow things down. The formula used to estimate the maximum speed that may be obtained is as follows: Throughput = TCP Window Size / Latency. Upon attempting to download many files at once from different places, a speed closer to the full 10G speed may be experienced. The formula being referenced describes how to determine the maximum theoretical throughput for a TCP session given a specific TCP window size and round-trip latency. In essence, the throughput of a TCP stream can be limited by its window size and the round-trip time (RTT, or latency) between the sender and receiver. The following is the formula. Where:   It is important to note that in real-world scenarios, achieving this maximum throughput can be influenced by various factors like network congestion, packet loss, and other variables. For a formal reference on this topic, consider reviewing RFC 793, which defines the Transmission Control Protocol (TCP). For a deeper dive into TCP performance and its relationship with window size and RTT, 'Computer Networks' by Andrew S. Tanenbaum is an excellent resource. Below are examples of how latency can interfere with the final capacity of a 10Gbps link. Example 1: Performance of a 10 Gbps Link with High Latency Given: Link Capacity: 10 Gbps TCP Window Size: Let's assume $65,535$ bytes (or $524,280$ bits since there are 8 bits in a byte) Round Trip Time (Latency): 100 ms (or 0.1 seconds) Using the formula:   Example 2: Performance of a 10 Gbps Link with Low Latency Given: Link Capacity: 10 Gbps. TCP Window Size: Let's assume $65,535$ bytes (or $524,280$ bits since there are 8 bits in a byte) Round Trip Time (Latency): 1 ms (or 0.001 seconds). Using the formula: How to Get the Most from the 10G Link: Check The Equipment: Make sure everything, from the computer to the cables and devices in between, supports 10G speeds.If even one part doesn't, it can slow everything down. Know The Hardware's Limits: Some things that say they can handle 10G speeds might not be able to, especially if the task is complicated. Try Many Downloads at Once: If many files are downloaded from different places at the same time, speeds closer to 10G may be experienced. Use Traffic Generators: These tools can push the link to its limits. They will aid in seeing whether 10G is really possible. Conclusion: Even if a super-fast 10G link is present, other factors such as the time it takes for data to travel (latency) can change the actual speed. If data has to travel a long way, it can slow down the download speed a lot. But if it only has to travel a short distance, a speed very close to the full 10G may be obtained. The important thing to remember is that many factors play a part in internet speed. It is always a good idea to test and understand these factors to get the best results.