What is non-blocking IO, how does it work, and how can you increase your throughput by a few orders of magnitude. What is the C10K problem, and why can't we just add more threads? I will also speak about when it's worth the extra complexity price, and how can you get there relatively easily.