[Costless] - 12~16 page 대본

12 Page

 

 

Second factor that affects the price is -> running edge or cloud. 

 

Running on edge device costs around $0.18.

 

----

 

13 page

 

Next, we want to run a function on the cloud and It needs intermediate data from edge device, we use storage service S3. We use API to retrieve data from S3 to run the lambda function on the cloud. This process incurs some transmission time. So we have to consider this price when communicating with the cloud by saving the intermediate data.

 

---- 

 

14 page

 

The last factor is to consider is memory allocation. If we double the memory size the CPU capacity doubles, but this doesn’t mean the function’s latency will reduce by half. Because we don’t if the application utilized CPU fully. Memory allocation is complicated problem, so for the rest of the paper, we focus on only one edge and cloud configuration. There is some information about exploring function fusion and placement with different memory configurations in the last part of the evaluation. We can take a look at it and discuss about this. 

 

 ----

 

15 page

 

 

Now that we know the factors that affect the price we can formulate the problem.

 

Simply, it’s about minimizing the price model when the Execution time model is less than the threshold we set.

 

----

 

16 page

Price model’s equation is as follows: