1: \begin{abstract}

2: Edge computing has been proposed to cope with the challenging

3: requirements of future applications, like mobile augmented reality,

4: since it shortens significantly the distance, hence the latency,

5: between the end users and the processing servers.

6: %

7: On the other hand, serverless computing is emerging among cloud

8: technologies to respond to the need of highly scalable event-driven

9: execution of stateless tasks.

10: %

11: In this paper, we first investigate the convergence of the two to

12: enable very low-latency execution of short-lived stateless tasks,

13: whose computation is offloaded from the user terminal to servers

14: hosted by or close to edge devices.

15: %

16: \added{%

17: We tackle in particular the research challenge of selecting the

18: best executor, based on real-time measurements and simple, yet

19: effective, prediction algorithms.

20: %

21: Second, we propose a performance evaluation framework specifically

22: designed for an accurate assessment of algorithms and protocols in

23: edge computing environments, where the nodes may have very heterogeneous

24: networking and processing capabilities.

25: %

26: The proposed framework relies on the use of real components on

27: lightweight virtualization mixed with simulated computation and is

28: well-suited to the analysis of several applications and network

29: environments.

30: }%

31: %

32: Using our framework, we evaluate our proposed architecture and

33: algorithms in small- and large-scale edge computing scenarios, showing

34: that our solution achieves similar or better delay performance than

35: a centralized solution, with far less network utilization.

36: \end{abstract}

37: