Problem: Cloud computing resources are usually centralized into large data centers

  1. data exchanged between end-devices and cloud resources may experience high latency
  2. typically located far away from the majority of end-users
  3. the increasing amount of data being transferred also demands high-bandwidth connections to the cloud.

Solution: Placing computing resources in close proximity to the devices generating the data reduces communication latency

The concept of bringing content closer to end-users is not new.

  • CDN
  • ICN

What’s new from edge Cloud Computing?

However,

  • the CDN and ICN paradigms are limited to non-interactive content.
  • edge computing servers also provide computational capabilities and can host interactive applications that support user mobility.
  • an edge computing platform can relieve privacy concerns as the data generated from IoT devices are stored and processed within nodes in the edge network.
  • data can be pre-processed to remove private information before being sent to the cloud
  • offloading computation to resources closer to the users (and data centers that are not far away) can help reduce the energy consumption at the end-devices

How to build Edge computing platforms?

Three possible ways:

  1. resource-rich servers deployed close to the end-devices
  2. leverages resources from heterogeneous nodes at the edge
  3. federation of resources at the edge and centralized data centers

**
Way 1:

  • A cloudlet is described as a “data center in a box” that offers a cluster of multicore computing capacity, storage and wireless LAN connectivity towards the edge.
  • Video and sensor data collected from users through Google Glass are processed on the cloudlet to provide real-time assistance.
  • Mobile-edge computing

Way 2:

  • heterogeneous mobile devices opportunistically share their resources to deliver services and applications.

Way 3:

  • Federation of clouds is also a key aspect of the FUSION

GigaSight - 3-tier structure

This system allows users to perform content-based searches on the captured videos

  1. Denatured: the process of removing privacy sensitive information from the video
    1. Each user is associated with a VM, which performs denaturing according to user-specific settings
    2. A separate VM on the cloudlet indexes the videos by analyzing the denatured video and generating tags for each frame
    3. Cloudlets are used to store the videos and only meta-data (including the content tags) about these videos is stored on a database on the cloud.

      the authors conclude that edge cloudlets perform better than a single larger metropolitan cloudlet

tactile Internet

the tight integration of digital data and physical environments through real-time wireless communications

  • combine augmented reality and sensor information such as user location: Pokémon Go and Ingress
  • the application needs to create a virtual environment and adapt it according to the user inputs.
  • cloud resources can be located far away from the user, thus introducing significant delays. While they would be acceptable for non-interactive applications, these delays cannot be tolerated in the considered use cases, as they would result in a poor user experience

Indeed, edge computing has been proposed as the reference paradigm for the tactile Internet.

cloudlet and mobile edge computing platforms described earlier, only the actual applications will decide which solution is the most suitable to satisfy their requirements.