Context-aware caching scheme (CACS) for real-time health monitoring systems

1
Context-aware caching scheme (CACS) for
real-time health monitoring systems

• Aarti Munjal, Aravind Kalavagattu
• Arizona State University
• CSE 535 Mobile Computing
• Project presentation

2
Problem Statement

• Formally,
• To design a context aware caching scheme for
  real-time health monitoring of patients in a
  community setting. The system has to be scalable,
  with minimum time-lag for information discovery,
  robust in handling the network traffic and ensure
  high amount of accuracy.

3
Tasks involved..

• Topology
• CACS
• Employing Context awareness
• Search Update
• Analysis
• System Design
• Simulation/Implementation

4
Topology

• To make it scalable, we adopt a multi-tier
  architecture
• PDAs at leaves
• Patient with sensors
• Server as root
• Doctor
• Hubs sit at intermediate levels

Server
H8
H7
H3
H2
H1
H4
H5
H6
P5
P6
P7
P9
P10
P11
P12
P8
P4
P2
P1
P3
Figure 6 Multi-tier architecture
5
Context Aware Caching Scheme (CACS)

• Aim
• To ensure data availability even in case of
  disconnections and failures.
• Cache Resolution Data request aimed to be solved
  at the highest level possible.
• Query should be served by the top-level hubs with
  out going to the patient PDAs always.
• Cache Management Decision regarding purging less
  important data to make room for the critical
  information.
• In case of overflow in traffic, critical data
  needs to be stored as compared to less critical
  ones

6
Data Structures Used for Caching

• Path
• Hub sequence is stored to reach the patient PDA

7
Employing Context-awareness

• Context
• Environmental Conditions
• Patients health vulnerabilities
• Range of each sensor values (if it is within the
  safe limit or not)
• Eg Temperature of 98-99 F is safe, but we need
  attention if it crosses beyond
• Context parameters
• Priority f(v,m,e) used for admission
  control at hub level
• TTL g(v,m,e) used to make sure the
  values get
• refreshed
  timely
• where,
• Defined at the PDA level.
• v variance of data values
• m mobility of PDAs (patients)
• e environmental factor

8
Search and Update

• Search
• Request (pid,sid,path) from Server level
  reaches the last hub in path.
• do
• go to immediate parent hub
• broadcast to the child hubs
• while (entry_not_found)
• Update
• Register Packet from pda (with ret 1) is
  sent to hub. Hub adds the entry in its data
  table, appends its own id in path field and
  forwards the packet to next-level hub.
• Packet is forwarded until it reaches the server.

Server
H8
H7
H3
H2
H1
H4
H5
H6
P5
P6
P7
P9
P10
P11
P12
P8
P4
P2
P1
P3
Figure Multi-tiers
9
Analysis Comparison with Simple Caching
Scheme

• Performance Measures
• Average cost for search and update
• Request Satisfaction factor
• Parameters for Evaluation
• N Total No. of Nodes
• n No. of levels
• r Branching Factor
• p Probability that each hub has data
• ?r Request Generation Rate
• ? PDA Mobility Rate
• c Average No. of levels for broadcast
• a Probability that path to PDA is known

10
Analysis Contd

• Average Cost Function for CACS
• C ?c (1-p)n-1an(1-a)(c r-1)
  (1-(1-p)n)(n-1)/2 ?m(n)
• Average Cost Function for Simple Caching Scheme
• C ?c (1-p)an (1-a)(N-1)/2 ?m(n)

11
Analysis contd

• Search cost ?c (1-p)n-1an(1-a)(c r-1)
  (1-(1-p)n)(n-1)/2
• (1-p)n-1 - probability that none of IM hubs has
  data
• n - no of hops travelled if path to pda
  is known (probability a)
• o/w we search path by broadcasting (probabilty
  (1-a))
• 1- (1-p)n - probability that at least one of
  the IM hubs has data
• (n-1)/2 - of hops to travel on an average
• Update cost ?m(n)

12
Results
No of nodes (N) Average Cost (simple caching scheme) Average Cost (CACS)
15 36 31
31 55 37
63 85 44
127 140 52
255 242 59.7
511 441 67.4
1023 831 75.03
13
Simulation

• Simulation tool Network Simulator (Ns2)
• Simulation in Ns2 is agent-based, where agents
  communicate with each other through
  message-passing.
• Three Agents
• Pda Agent (lowest level)
• Hub Agent (Intermediate levels)
• Server Agent (Highest level)
• Two types of packets
• Request packet
• Data packet

14
Simulation contd

• Storage
• Tables
• Communication
• Packet Structure
• Timer Events (TTL-based)
• PDA Refreshes data and sends to hub
• Hub Request sent to PDA
• Timer Values hashed using ltpid,sidgt pairs

15
Tasks accomplished

• Mathematical analysis of CACS
• Its comparison with the traditional simple
  caching scheme
• Proved CACS performs much better than the
  traditional caching scheme.
• Implemented the design of system for simulation
• Created three agents
• Data Structures used to store data at each level
• Timers to refresh TTL
• Incorporating Context-awareness using rules
• Data admission control based on priority

16
Future Work

• Request Satisfaction Factor (rs)
• Number of requests satisfied/ Number of requests
  generated
• Simple Caching Scheme
• Data stored using FCFS.
• CACS
• Critical data always given preference.
• rs for CACS gt rs for Simple Caching Scheme
• Due to the context-aware caching stores most
  frequently asked or critical data all the time.
• Simulating the system for a large community of
  patients for testing and validating the
  mathematical analysis.

17
References

• Krishna Venkatasubramanian, Guofeng Deng, Tridib
  Mukherjee, John Quintero, Valliappan Annamalai
  and S. K. S. Gupta, Ayushman A Wireless Sensor
  Network Based Health Monitoring Infrastructure
  and Testbed, IEEE International Conference on
  Distributed Computing in Sensor Systems (DCOSS),
  2005
• Y. Du and S. K. S. Gupta, COOP - A cooperative
  caching service in MANETs, In Proc. of ICAS-ICNS
  2005. Joint International Conference on, Tahiti,
  French Polynesia, pp 58-63, Oct. 23-28, 2005
• Anurag Kahol, Sumit Khurana, Sandeep K.S. Gupta,
  and Pradip K. Srimani, A Strategy to Manage Cache
  Consistency in a Disconnected Distributed
  Environment, IEEE Transactions On Parallel And
  Distributed Systems, VOL. 12, NO. 7, JULY 2001
• A. Skordylis, N. Trigoni and A. Guitton, A Study
  of Approximate Data Management Techniques for
  Sensor Networks, Proc of the 4th Workshop on
  Intelligent Solutions in Embedded Systems, 2006.
• Guanling Chen and David Kotz. A Survey of
  Context-Aware Mobile Computing Research,
  Department of Computer Science, Dartmouth
  College, Dartmouth Computer Science Technical
  Report TR2000-381