Today's hyper-connectivity of both people and things has led to an enormous jump in network traffic. Global IP traffic will increase nearly threefold over the next five years. It's not going to lessen, nor is it going to get slower. These factors create a need for high-speed networks to ensure service level and capacity. 100G network links are deployed by telecom networks and data centers, serving hundreds of thousands of users, so that these providers can keep pace with customer demand. This then creates the need to test and troubleshoot on the networks at 100G link speed.

Why network traffic replay is helpful
Quality and maturity are critical when deploying new solutions today because there are so many services and subscribers depending on stable data networks. Testing new network solutions or new versions of networking products or applications is a challenge because synthetic test cases do not disclose all potential issues. Replaying real production network traffic provides more realistic test scenarios. Moreover, the test coverage can be extended by replaying traffic from many different networks.

Good troubleshooting reduces the disruptions and potential down-time of services in case network issues arise. Network traffic replay provides a powerful tool for reproducing and analyzing issues in a lab environment, not affecting the production network. Identification of network issues requires precise replay, throughput-wise and timing-wise. For example, traffic micro-burst can only be recreated in a lab environment if the solution supports full throughput and precise timing of packets replayed.

Approaches to replay
It's possible today to purchase a complete solution for capture and replay of 100G network traffic. In some cases, these products match the actual need, but in other cases the solutions are much too expensive, have too many features or cannot be customized to special use cases.

There is another option, though. It is also possible to build your own replay solution, using a COTS server, a standard Linux OS and a 100G smart NIC. The COTS server has all the horsepower and I/O performance needed to run 100G replay. Linux OS enables use of open source software or easy application development and feature customization. The smart NIC ensures full throughput transmit for any packet size, with precise transmit timing, with some smart NICs having one-nanosecond time resolution.

DIY 100G replay solution
Certain qualities are needed from the raw materials that will comprise a high-performing 100G replay solution. For starters, to obtain maximum replay performance, it is recommended to have maximum memory bandwidth available in the server. The server should be populated with the maximum number of DDR4 memory blocks to allow full use of all the memory channels. Another recommendation is to use dual rank DDR4 or memory blocks running at least 2133 MHz or higher. To replay large files, the packet store output performance limits the maximum average replay speed. Fast Solid State Disks should be considered for high-speed replay.

To be effective in this set-up, the smart NIC must support a 100G network interface with line-rate capture and transmit for any packet size. It also needs high-precision time stamping of received traffic, high-precision transmit timing, an efficient buffer system and an optimized DMA scheme over the PCIe interface.

It's possible to get capture and replay applications that are open source, and some smart NIC vendors include tools for capture and replay in their software suite. The application should use server memory for buffering traffic from packet store before transmission. This is to compensate for the difference between 100G peak network packet rate and maximum read rate from the packet store.

Choosing what's best
Network traffic is growing at such an accelerated rate that high-speed networks are becoming a necessity. Organizations that depend on 100G network links to meet service level agreements and capacity needs must test and troubleshoot those links. This is where network traffic replay becomes so helpful; it enables testing outside the production network. Complete capture and replay solutions are available for purchase, or you can choose to build your own. You may find that a custom solution will reduce costs while still delivering on promised network performance.