ExpertTCP™ - TCP Throughput Testing (RFC 6349) on PacketExpert™ 100G

PacketExpert™ 100G - SmartNIC Specifications (Per Card)

Optics	2 x QSFP28 cages for 2 x 100 GbE, 2 x 50GbE, and 2 x 40 GbE Supports 2 x 25 GbE, 2 x 10 GbE, and 2 x 1 GbE with QSFP-to-SFP adapter
PCIe	PCIe Gen 3, 16 lanes
RAM	8 GBytes DDR4 SDRAM
1000Base-T Port	RJ45 for IEEE1588v2
Single-ended Coaxial I/O	SMA connector, 50 Ohms for External Clock Input/Output
Temperature Range	0 °C to 45 °C
Operating Humidity	20% to 80%
Storage	-10 to 60 °C
Oscillator Accuracy	+/- 4.6ppm

---

PacketExpert™ 100G Rack-mount Platforms

• Ideal for Lab environments that require centralized management of multiple servers and network devices
• Rack-mount units offer flexibility for scaling up or down as needed by adding or removing individual units
• The ExpertTCP™ application supports only on 2x1G rate with an aggregate TCP generation rate of up to 1 Gbps

---

PacketExpert™ 100G Portable Platforms

• Ideal for field engineers, military personnel, or researchers who need a powerful and portable computing solution in remote or rugged locations
• Suitable for environments where traditional desktops or laptops may be too fragile or lack necessary durability
• The ExpertTCP™ application supports only on 2x1G rate with an aggregate TCP generation rate of up to 1 Gbps

Test Configurations

Functional Procedure

The ExpertTCP™ Test Configuration on the PacketExpert™ 100G platform allows users to perform structured RFC 6349-based TCP throughput validation between two PacketExpert™ units configured as client and server. The configuration workflow automatically guides users through the recommended RFC 6349 methodology to ensure accurate TCP performance measurement.

ExpertTCP™ specifies the TCP Throughput test to be conducted in 3 steps:

• Path MTU Discovery
• Determine Baseline RTT
• Conduct TCP Throughput test

The application supports all 3 tests in a seamless one touch way. User can configure the parameters and simply start the test. ExpertTCP™ will run through all the three tests and report the results.

Path MTU Discovery – As specified in RFC 6349, the first step is to determine the Path Maximum Transmission Unit (MTU) supported across the network path. The TCP throughput test is then performed using this Path MTU value to prevent packet fragmentation, which can otherwise impact TCP performance and lead to inaccurate test results.

ExpertTCP™ determines the Path MTU separately for both upstream and downstream directions. It follows a method similar to the approach defined in RFC 4821 (Packetization Layer Path MTU Discovery), using TCP-based probing instead of ICMP. The discovered Path MTU is then used to determine the appropriate Maximum Segment Size (MSS) for the TCP throughput test.

Determine Baseline RTT – This step measures the inherent, non-congested Round-Trip Time (RTT) of the end-to-end network path. RTT represents the time taken for a TCP data packet to reach the destination and for the corresponding acknowledgment (ACK) packet to return to the sender.

RTT is an important metric in TCP, as it determines how much data can be transmitted before an acknowledgment (ACK) is received. This measurement is used to estimate the optimal TCP Receive Window (RWND) required for the subsequent Throughput test.

ExpertTCP™ performs the Baseline RTT test separately for each direction and automatically calculates the optimum TCP RWND based on the measured results. The corresponding Bandwidth-Delay Product (BDP) and RWND values are displayed for user reference.

Conduct TCP Throughput Test – In this step, TCP throughput testing is performed using single or multiple TCP connections to determine the achievable TCP throughput. The TCP Receive Window (RWND) calculated during the Baseline RTT test is applied for the throughput measurement. When multiple TCP connections are configured, the available RWND is distributed across the connections. The system supports configuration of up to 16 TCP connections.

Statistics and Results

The Statistics and Results section provides visibility into TCP performance measured during the ExpertTCP™ RFC 6349 throughput test. Results are displayed separately for upstream (Client → Server) and downstream (Server → Client) directions to help evaluate end-to-end TCP performance across the network path.

• Path MTU – Maximum transmission unit supported across the network path without fragmentation
• Baseline RTT – Inherent round-trip delay of the network path prior to throughput testing
• Bandwidth-Delay Product (BDP) – Optimal amount of data that can be in transit on the network path
• TCP Receive Window (RWND) – Calculated window size required for efficient TCP throughput
• TCP Throughput Efficiency – Effectiveness of bandwidth utilization during the throughput test

These measurements help determine the maximum achievable TCP throughput and identify performance limitations caused by latency, packet loss, retransmissions, or buffer constraints.

In addition, the system displays per-connection throughput statistics including average, minimum, and maximum throughput, along with RTT measurements, retransmission counts, and transfer duration. Detailed transmission statistics such as transmitted and received frames and bytes are also provided for each configured TCP connection, allowing users to evaluate connection-level behavior during the throughput test.

These statistics help network operators, service providers, and test engineers verify whether Ethernet/IP networks that meet Layer 2/Layer 3 performance benchmarks can also sustain expected application-level TCP throughput performance required for SLA validation, deployment verification, and troubleshooting scenarios.