Reading notes by Yu Guan: Low-power Internet of Things with NDN & Cooperative Caching

2017-11-11

Posted by 关宇

This is a resubmitted paper of 2016 GLOBECOM “A Named Data Network Approach to Energy Efficiency in IoT ”

The propose of this paper:

energy efficiency in the IoT

At present, there are several methods to save energy in IoT, they can be considered as several design space. So an energy-saving architecture can be regarded as the combination of the following 4 methods.

  1. Energy efficient hardware with micro-controller and radio consuming energy in mW range and ultra-efficient sleep modes in nW range. Energy harvesting techniques may also be applicable in some cases.
  2. Radio duty-cycling (RDC) at the MAC layer. They achieve low power by minimizing idle listening.
  3. Less chatty network layer protocols avoid communication in broadcast/multicast as the 6loWPAN protocols that adapt IPv6 to the IoT.
  4. Centralized content caching in the cloud or on a proxy, e.g. CoAP / HTTP caching.

In this paper, authors mainly focus on the last design space——how to do energy saving through caching technology.

 

Standard approach and drawbacks

Standard approach: With centralized content caching in place, content availability is preserved by a proxy or the cloud, while IoT devices sleep a large part of the time.

Drawbacks:

  1. when the local network gathers a large number of nodes, explicit synchronization and coordination of RDC with a MAC layer based on TSCH becomes impractical
  2. connectivity with the designated gateway/proxy is intermittent, and centralized caching of IoT content fails.

 

So, in this paper, authors propose in-network caching in IoT.

The purpose of this paper is maximizing the tradeoff of availability and energy saving.

Objective function:

max{alpha*A(c, p) + beta*E(c, p)}

A: availability

E: energy

constants:

L: constant. The freshness of requested data. L = x means that the requested data must be the data in the latest x versions. Else this data is regarded unavailable.

variations:

c: caching strategy. In this paper, authors tests 3 strategies: no cache, random caching and MDMR.

p: sleeping probability. Each device has p probability in sleep and 1 – p awake.

 

Conclusion

In-network caching can provide high availability and energy efficiency compared with no in-network caching.

 

Some ideas about this work

  1. Different sensors should have different behaviors. Simply configure all sensors’ sleep ratio to p doesn’t make sense. We need to acknowledge the behavior of different kinds of sensors and formulate the objective function in detail.
  2. Although coordinate sleep is compared to this solution, this solution doesn’t apply coordinate sleep. I think coordinate sleep strategy is also useful in energy-aware caching strategy.
  3. This work is a pure mathematical problem. I think it is far way from actual usage.