DataCentric Storage in Sensornets Submitted to Sigcomm 2002

1
Data-Centric Storage in SensornetsSubmitted to
Sigcomm 2002

• Authors Sylvia Ratnasamy et al.
• ICIR, UCLA, UC-Berkeley
• Presenter Shang-Chieh Wu (meou_at_cs)

2
Sensornets

• Sensornets are large-scale distributed sensing
  networks comprised of many small sensing devices
  equipped with memory, processors, and short-range
  radio
• Monitor temperature in an active volcano, keep
  track of the seismic waves in an earthquake
  region, or even save rare and endangered species.

3
Sensornets

• Scalable, self-organizing and energy-efficient
  data dissemination algorithm required.
• Data-centric routing algorithm.
• Data-centric storage.

4
Terminology

• Observations The low-level output from the
  sensors.
• Events Certain constellations of low-level
  observations.
• Task The code running on each sensor.
• Query Elicit the event information from the
  sensornet.

5
External, Local, and Data-Centric Storage

• External Storage(ES) On detection of events, the
  relevant data is sent to external storage.
• Local Storage (LS) Event information is stored
  locally.
• Data-Centric Storage (DS) After an event is
  detected, the data is stored by name within the
  sensornets.

6
ES, LS, DS

• Assume n nodes totally and the diameter of the
  sensornet is O(n0.5)

Communication Costs per each
7
ES, LS, DS

• Dtotal The total number of events detected.
• Q the number of event types for which queries
  are issued.
• Dq the number of events detected for the types
  of events queried for.
• No more than one query for each event type, so
  there are Q queries in total.
• Assume hotspot occurs on packets sending to the
  access point.

8
ES, LS, DS
Approximate Communication Costs

• DS is preferable in
• Sensornet is large
• Dtotal gtgt maxQ Dq (??)
• Summarises are used.

9
Data-centric storage

• Based
• GPSR geographic routing algorithm.
• Efficient peer-to-peer lookups system.
• Relevant data is stored, by name, at nodes within
  the sensornet. All data with the same general
  name will be stored at the same sensornet node.
• As an alternative method.

10
GPSR

• Strongly geographic routing algorithm.
• Each mode determines its own geographic
  coordinates and periodically announces its
  address and coordinates to its neighbors.
• Each mode is aware of the identifications and
  positions of all its neighbors.
• Special rule to handle dead-ends.

11
GPSR
12
GPSR
Right-hand rule Each node to receive a packet
forwards the packet on the next link
counterclockwise about itself from the ingress
link.
13
Distributed Hash-Table

• Given an event name, we hash the name into a key.
  The key is a location somewhere in the
  sensornets.
• Put(key, value) command sends a packet with the
  given payload into the sensornet routed towards
  the location key.
• MITs GLS is a good candidate.
• Some extra concepts are proposed to handle
  failure, mobility, and scalability.

14
Conclusion

• Storages are involved in sensornets, another
  option.
• A detail evaluation for transmission cost from
  collecting data to answering queries.

15
My opinions

• Database approaches for sensornets.
• No serious contribution in real key
  issue.(routing and hashing)
• Oversimplify the difficulty of summary events.
• Lengthy introduction for sensornet.
• The cost of failure for storages is neglected, On
  purpose ?
• Data consistence in storages ?