Testing for 10G throughput

Question

I'm currently facing a throughput limitation in my small core network setup, and I'm seeking advice on how to optimize it for higher speeds. Here's the current setup: Equipment: Two small core setups, each with Switches, ASAs (Adaptive Security Appliances), WLC (Wireless LAN Controller), etc. The Core Switch is connected to the ASA on both kits. Both switches are interconnected. Each kit has a laptop connected to the switch, behind the ASA, for testing.

Issue: I'm aiming for 10G throughput, but I'm only achieving 1G due to port channel configurations. (since the throughput of a port channel is typically limited to the bandwidth of the individual links in the bundle.) I understand that having more than one IP on each side might help improve the throughput.

Questions: How can I reconfigure/change my setup to maximize throughput? Is there software available that can simulate multiple clients to test network performance?

Any guidance, tips, or recommended tools/software for testing and optimization would be greatly appreciated. Thank you in advance for your assistance!

Answer 1

I'm only achieving 1G due to port channel configurations. (since the throughput of a port channel is typically limited to the bandwidth of the individual links in the bundle.)

The throughput for individual flows is limited to that of the individual links in the LAG. The throughput of the entire bundle with a decent number of flows should be much higher. However, expecting 10G-ish throughput across eight aggregated 1G links isn't realistic - you'll need 10G ports for that.

I understand that having more than one IP on each side might help improve the throughput.

That depends on the workloads and on the switches' traffic distribution policy (may or may not be configurable). Multiple IP addresses may help with some application protocols (off topic here) to create a sufficient number of flows, but it's no guarantee to get a decent distribution. Most often, multiple IP addresses provide better distribution over individual L3 ports instead of an aggregated group though.

Traffic distribution policies are usually based on a hash over source&destination MAC (worst), or source&destination IP (better), or source&destination IP:L4proto:port (best), sometimes in conjunction with VLAN tags. Some switches have fixed policies, some are configurable.

How can I reconfigure/change my setup to maximize throughput? Is there software available that can simulate multiple clients to test network performance?

Basically, you need to analyze your workloads and evaluate which switch/config suits those loads. And make sure that the heavy workloads are generally L2-/L3-switched rather than running over the ASA, wherever possible.

And of course, you need to make sure that your gear can actually handle 10G workloads - storage, databases (all off topic here) may require optimization or upgrades.