Adaptive Protocols for Informatin Dissemination in Wireless Sensor Networks

Summary. The authors present SPIN, a protocol for info dissemination in computatinally and power challenged wireless sensor networks that features:

• meta-data negotiation prior to data exchange to ensure that the latter is necessary and desired..
• Power resource awareness.

More detail.

• improve accuracy of information obtained via collaboration among sensor nodes and online processing of that information.
• when networked, can focus on critical events (such as an intruder entering a secure area).
• sensors can aggregate data to provide a rich view of an environment.
• provide fault tolerance by sensing that a sensor has failed and covering its domain.
• can improve remote access to data by providing sink nodes that are connected to external networks.

Challenges of wireless sensor network devices:

• Energy: must minimize computatin and communication.
• Computation: some sensors have limited computation capability.
• Communications: bandwidth is often limited (on the order of hundreds of Kbps)

Challenges of wireless sensor network protocols:

• Implosion: conventional protocols can multiple copies of data to arive at a single node.
• Overlap: sensor domains might overlap leading to duplicate information dissemination.
• Resource blindness: conventional protols do not take into account available power.

SPIN overcomes the three protocol difficiencies above using (1) meta-date negotiation prior to data transmissin to ensure that the receiver desires and needs the data and (2) resource-adaptation to allow the device to be more prudent about sending data when power supplies are low.

The SPIN-1 Protocol:

• Three packet types:
  1. ADV: new data advertisement.
  2. REQ: request for data.
  3. DATA
• The data transmissions are receiver-driven since the sender will only send data if a receiver requests it.
• Data can be aggregated and advertised aggregately.

SPIN-2 Protocol:

• SPIN-1 + resource awareness
• When the app detects low power, it will not proceed with a round of the protocol unless it knows it has enough power to complete the round with all of its neighbors without falling below a minimum power threshold.

Extended ns to understand SPIN. Simulated several conventional protocols and SPIN protocols:

• flood: upon receipt of new data, send to all neighbors
• gossip: upon receipt of new data, send to only one neighbor; can send to sender (ensures that if networks are partitioned by a single node, both sides of the partition will get the data).
• ideal: no duplicate data, best-case convergence; simulated using the SPIN data paths after SPIN had performed negotiations (they send no duplicate data packets).
• SPIN-1
• SPIN-2

Results:

• In terms of time, SPIN-1 achieves comparable results to classic flooding protocols and in some cases outperforms flooding (when initial data is not unique).
• In terms of energy, SPIN-1 uses about 25% as much energy as classic flooding. SPIN-2 is able to distribute 60% more data per unit energy than flooding.
• In all experiments, SPIN-1 and SPIN-2 outperformed gossiping.

Questions / Comments

• I assumed that they would package a device's entire namespace into one advertisement. If so, that would take considerable computational energy and, in many situatinos, more than 16 bytes. If this is the case, than the fact some of the power savings realized by confirming that data transmission are desired by the recipient might not be realized.

  Or is my assumption incorrect and they are going to send out one ADV per new piece of data that doesn't encapsulate the device's namespace?

• In the experiments where power was limited to 1.6 Joules, one of the protocols uses 1.8 Joules.
• The paper does acknowledge that lossy networks must be addressed, but it was disappointing that it wasn't addressed in this paper considering that it's domain is wireless networks where losses are more common than in wired networks. Note: Apparently this is addressed to some extent in a later paper.