Expert Answer
Anonymous
The most important tenet is to understand the needs of the audience. For example, I recently partnered with the Product team to improve the elastic cluster experience to support bursts in traffic of up to 10x past peaks as opposed to 4x that it was. The Elastic team had broken down the ask into engineering requirements and came to the backend with the need to improve the rate of resizing by 30% in the short term and 4x in the long term along side asks to increase our baseline capacity footprint by ~2.5x worldwide. They'd also taken on significant scaling projects on their Elastic Workflow Engine and had finally given Product a 9-month roadmap before feature launch. When I looked at the problem statement I saw a path to providing this feature trivially to the customer as long as we're able to charge them differently for 10x support instead of 4x: instead of having Elastic resources to be twice as big as their peak load, make them 5 times as big from the get go and pass along the extra cost to the customer. We can do this with fairly straightforward tweaks to the Elastic engine and doesn't involve any work on the backend. The Elastic team EM and Staff engineers had suggested the "right", "most efficient" solution to Product. This allowed incremental delivery, tune-ability (3x, 4x, 5x support) etc. But, Product was interested in ensuring key customers don't walk away from Elastic clusters today. These customers didn't expect to pay the same amount for the increasing performance. Explaining the constraints of the system and the factors that played into cost modeling in simple terms allowed us to favor my proposal. The project was delivered within 30 days of the pivot in private beta and in 2 months it was available worldwide behind an allowlist.