Alice and the head trader, Dave, kick off the meeting by summarizing the production problem. You digested a lot of insight into the problem from the meeting. You learned that currently, there is high market volatility and the rapid swings in prices amplify opportunities to generate profitable trades. Unfortunately for MVT, in recent weeks, the high volatility has brought with it unseen levels of order volume. Traders have been flooding the markets with limit orders and cancelations to react to the quickly changing price action. The MVT order book is certified via load testing to handle a maximum of 15,000 orders per second (OPS) with a 99^th percentile latency of 10 milliseconds (ms). The current market conditions are producing sustained levels of 45,000 OPS, which is destroying tail-end order book latencies. In production, the deployed version of the order book is now producing 99^th percentile latencies of up to 80 ms and maximum latencies reaching 200 ms. In the trading business, a slow reaction can quickly turn a profitable trade into a sizable loss. This is exactly what has been happening at MVT. Typically, in times of volatility, MVT is able to generate healthy returns, but in recent weeks, there have been staggering losses. Our goal is to apply the techniques that we learned in Chapter 1, The Road to Performance, to make inroads on the performance woes.

The traders at MVT are looking for a quick performance win to take advantage of the current market environment. The traders believe that the market volatility will persist for another week before subsiding. Once the volatility disappears, so do money-making opportunities. Therefore, it's been stressed to the engineering team that an incremental reduction in 99^th percentile latency to 40 ms should halt trading strategy losses and actually produce small profits. Once the volatility subsides, more in-depth and extensive performance improvements are welcomed. For now, the clock is ticking, and we need to find a way to stop the losses by improving performance incrementally.