.png){kind=icon}
.png){kind=icon}
Return
An adaptive function placement in serverless computing
.png){kind=icon}
PRE
AI
DOI：10.1007/s10586-021-03506-x.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Serverless computing is a new fine-grained computing and deploying paradigm which combined terminal devices, edge nodes and cloud data center into a complete computing system. Stateless function is the critical factor to implement serverless computing. For optimizing the function execution, function placement is becoming the urgent problem to be solved in the serverless computing. To specify what is function placement in serverless computing, an adaptive function placement framework is presented. For prejudging the trend of function placement, positive or negative properties of event are classed. Based on the positive or negative impact on the function placement decision and function placement trend, mathematical model of function placement is presented;
the first time. For specifying the adaptive function placement in serverless computing, An adaptive function placement model is;
mulated based on the Markov Decision Process (MDP). In the model, function states space, placement rate, placement probability, decision time, action set, cost and criterion are presented;
the first time to specify the computation of the function's MDP.The presented algorithm decides function execution in terminal devices,local edge nodes or the function execution in the remote available cloud servers in a real time and adaptive way. The presented algorithm also permits dynamically allocating functions to minimize the Dec(t) while keeping performance satisfaction. Evaluation and experiment show the presented adaptive function placement algorithm can get the satisfied requirements of different indexes.
Keywords:
Serverless computing
Stateless function
Function placement
Agent
Journal
C
IF:
4.1
Papers:
5.0K
Citations:
7.5K
Organization
No organization information available