Steps
- Create a baseline on one stable profile: real delay, sustained throughput, CPU use and failure rate on a fixed network.
- Compare another server or network first. A large change means the bottleneck is not an advanced client option.
- Test one variable such as core, Mux, TUN or MTU while keeping target and time comparable.
- Keep only repeatable gains and revert neutral changes. Retest after core upgrades because defaults evolve.
Why this matters
Without a baseline, you cannot separate improvement from ordinary network variance. Controlled A/B tests reveal whether a setting fits your path.
How to verify
Several runs show lower error rate or better throughput without abnormal CPU, memory use or application breakage.
Things to watch
Mux does not always reduce delay. An MTU that is too small wastes capacity; one too large can cause fragmentation or black holes.