[Costless] - Revised one with transcript - p 6~10

 6 Page

The idea this paper proposes is first formulating the problem of optimizing the price and execution time as 2 models. First is the price model for AWS Lambda, and execution time model for workflow functions based on the functions costs and transition costs.

 

Secondly, it proposes possible function fusion and placement in a structure which is a cost graph. we formulate the problem as a constrained shortest path.

 

7 page & 8 page

When executing lambda, we have to consider 4 factors. 

1. The number of times each function is executed per month

2. The memory allocated to the function

3. The duration of how long the function runs.

4.  The price per 1 GB of memory and 1 second of execution

9 page 

 

 

For example the first function face detection. using 4 factors we mentioned, execution time * memory * duration * The price per 1 GB of memory and 1 second of execution will yield 16.67$

 

For the price of total functions, It would be adding price of each function which is 35$.

 

10 page

 

Beside each price of each lambda function. we have to consider the transition cost because this is workflow functions. If we add transition cost, it’s 185$ which shows transition cost could be more expensive than lambda function.