Tech Debts VS More Features

Smart folks are the ones who can effortlessly collect millions of users' personal pics freely to teach their AI.

Kudos to them.

Serverless: Built to be used less?

No need for a team to run servers. Serverless functions promise simplicity and cost-efficiency. You pay only when they work.

Sounds great, right? But here’s the catch: the more they run, the more you pay. If your Lambda is running constantly, it’s a red flag.

𝗧𝗿𝗮𝗻𝘀𝗮𝗰𝘁𝗶𝗼𝗻 𝗟𝗼𝗰𝗸𝗶𝗻𝗴 𝘃𝘀. 𝗔𝘀𝘆𝗻𝗰𝗵𝗿𝗼𝗻𝘆: 𝗧𝗵𝗲 𝗛𝗶𝗱𝗱𝗲𝗻 𝗖𝗼𝘀𝘁 𝗼𝗳 𝗖𝗼𝗻𝘁𝗿𝗼𝗹

Markets move in milliseconds. Every millisecond matters. Choosing wrong approach can cost you more than just latency. This isn’t just about database locks or async messaging. It’s a fundamental design choice.

𝗥𝗼𝗯𝘂𝘀𝘁 𝗼𝗿 𝗥𝗲𝘀𝗶𝗹𝗶𝗲𝗻𝘁? 𝗪𝗵𝗶𝗰𝗵 𝗢𝗻𝗲 𝗧𝗿𝘂𝗹𝘆 𝗠𝗮𝗸𝗲𝘀 𝗮 𝗦𝘆𝘀𝘁𝗲𝗺 𝗨𝗻𝗯𝗿𝗲𝗮𝗸𝗮𝗯𝗹𝗲?

→ 𝗥𝗼𝗯𝘂𝘀𝘁𝗻𝗲𝘀𝘀: can the system perform within predetermined expected boundaries.
→𝗥𝗲𝘀𝗶𝗹𝗶𝗲𝗻𝗰𝘆: can it adapt when the capacity to work is exceeded.

Or, Combine both - robustness to withstand and resiliency to recover?

𝗣𝗿𝗶𝗼𝗿𝗶𝘁𝘆 𝗤𝘂𝗲𝘂𝗲𝘀 𝗕𝗼𝘁𝘁𝗹𝗲𝗻𝗲𝗰𝗸𝘀: 𝗪𝗵𝗲𝗻 𝗽𝗿𝗶𝗼𝗿𝗶𝘁𝘆 𝗾𝘂𝗲𝘂𝗲 𝗹𝗼𝘀𝗶𝗻𝗴 𝗽𝗿𝗶𝗼𝗿𝗶𝘁𝘆

The idea from Chris G.'s Ph.D. dissertation. This shows how priority queues lose priority: bottlenecks can delay high-priority tasks and process lower-priority ones first.

𝗛𝗼𝘄 𝗱𝗼𝗲𝘀 𝗶𝘁 𝗹𝗼𝗼𝗸 𝗽𝗿𝗼𝗴𝗿𝗮𝗺𝗺𝗮𝘁𝗶𝗰𝗮𝗹𝗹𝘆: bit.ly/3R04KWn

𝗢𝗽𝘁𝗶𝗺𝗶𝘇𝗶𝗻𝗴 𝗰𝗼𝗱𝗲 𝗶𝘀𝗻'𝘁 𝗷𝘂𝘀𝘁 𝗮𝗯𝗼𝘂𝘁 𝘄𝗿𝗶𝘁𝗶𝗻𝗴 𝗶𝘁.

A key aspect of performance tuning is selecting the right data structure. I profiled stack implementations to optimize performance, and the results were mind blowing.

Here's what I found -

Software should be correct first, then clear, then fast. Factored is nice to have.

Your writing does the job it was designed for. It is written to be as easy to understand as possible. It does its job quickly.

It is structured into modular, reusable, well-organized, and maintainable components.