Cold Start: AWS Lambda vs Lolo Code

╔═════════════════════════════════════════╦══════════════════════════════════════════╗
║                AWS Lambda               ║                 Lolo Code                ║
╠═════════════════════════════════════════╬══════════════════════════════════════════╣
║ - Lambda is cost-efficient, infinitely  ║ - Less to no cold starts in Lolo as the  ║
║ scalable, serverless, but let's accept  ║ container is live always and will not be ║
║ the fact that, they do have some        ║ killed in case of inactivity.            ║
║ drawbacks. One such drawback is cold    ║                                          ║
║ start.  Cold starts in Lambda           ║ - Fundamentally Lolo code is different   ║
║ are frustrating.                        ║ from Lambda as the application/function  ║
║                                         ║ container will always be running.        ║
║ - Cold start prevents us to use Lambda  ║                                          ║
║ for customer facing applications and    ║ - In the free-tier plan, I ran a series  ║
║ mission critical applications.          ║ of experiments on the simple hello world ║
║                                         ║ API (stateless) and not sure about the   ║
║ - I ran a series of experiments on the  ║ container's memory and CPU restrictions. ║
║ simple hello world API, here are the    ║ Right now, Lolo allows the application   ║
║ results of the experiment. (Lambda      ║ deployment only in the EU (Europe)       ║
║ deployed in AWS EU)                     ║ region. I am testing these functions     ║
║                                         ║ from India and this attributes           ║
║ 1 minute later - 180 ms                 ║ to the delayed response time. Here are   ║
║ 15 minutes later - 750 ms               ║ the results of the experiment:           ║
║ 3 hours later - 1467 ms                 ║                                          ║
║ 10 hours later - 2060 ms                ║ 1 minute later - 172 ms                  ║
║                                         ║ 15 minutes later - 210 ms                ║
║                                         ║ 3 hours later - 602 ms                   ║
║ - Even though there are ways to resolve ║ 10 hours later - 890 ms                  ║
║ the cold start issue in Lambda but they ║                                          ║
║ are not foolproof. One such solution is ║ - The time represents the response time. ║
║ to keep invoking the Lambda in a fixed  ║ Although there is a slight invocation    ║
║ interval, so that the instance stays    ║ delay, but it is not as worse as Lambda. ║
║ live all the time, but this is not a    ║                                          ║
║ guaranteed approach.                    ║                                          ║
║                                         ║                                          ║
║ - On an average, the cold start         ║                                          ║
║ (invocation delay) is anywhere between  ║                                          ║
║ 2-7 seconds. The delay depends on       ║                                          ║
║ various factors such as underlying      ║                                          ║
║ runtime, allocated memory, etc.,        ║                                          ║
║                                         ║                                          ║
╚═════════════════════════════════════════╩══════════════════════════════════════════╝