Reliability Analysis in Net-Centric Computing

1
Reliability Analysis in Net-Centric Computing

• Supervisor Prof. Michael Lyu
• presented by Chong Ka Lung

2
Outline

• Introduction
• Net-Centric architecture
• Net-Centric configuration
• Cost/effective study
• Reliability study
• Future works

3
Introduction
4
What is net-centric computing?

• A new paradigm of client/server computing
• Applications are resided, managed, supported and
  100 executed on a server
• A multi-user operating system
• Server distributes the presentation of an
  application's interface to a client device

5
Benefits

• Single-point management
• Easy maintenance
• Physically and technically secure
• Predictable ownership costs
• Bandwidth-independent performance

6
Problems of Desktop PCs

• Fast obsolescence
• Difficult administration
• Expensive to maintain
• Time consuming for installation
• Virus attack

7
Solution - Thin client

• Replace desktop PC in net-centric environment
• A very small size of client operating system
• Reduced hardware costs
• Reduced maintenance and support costs
• Enhanced security

8
Size comparison
9
Net-Centric architecture configuration
10
Net-Centric architecture

• Net-centric server
• Platform-independent middleware
• supports heterogeneous platform of operating
  system
• Distributed component API, e.g. CORBA
• Wire/Wireless connection

11
(No Transcript)
12
Net-Centric configuration
13
Cost/effective study
14
Assumptions

• Thin client device is used instead of desktop PC
• At most 10, 30 and 50 clients are served
  concurrently
• same speed of processor type is used in desktop
  PC and in server

15
Comparison of number of hardware components (10
clients)
16
Cost comparison
17
Cost comparison (adding memory and upgrading
processor)
18
Reliability study
19
What is reliability?

• how long a system is continuously operational
• Probability that the system remains operational
  by time t given that it was started in perfect
  condition at time 0

20
Architecture
Server
Server
LAN Switch
Server
21
Fault tree Model
22
Parameter listing
23
Assumptions

• Components which will fail
• processor, memory, harddisk
• 3 software applications in server
• network connection on client and on server
• Independent behavior of components

24
Assumptions (2)

• In net-centric environment, thin client devices
  are assumed to be fail-free, while server will
  fail
• In traditional client/server environment, clients
  will fail as well as server

25
Assumptions (3)

• System is considered as failure
• processor fails
• memory fails
• harddisk fails
• 2 out of 3 software applications fail
• network connection fails

26
Control result
27
Optimistic failure rate
28
Fault tolerant effect
29
Comparison between net-centric and traditional
client/server
30
Future works

• Modeling on the non-independent behavior of
  components
• Building reusable component using CORBA
• Investigating the software fault tolerant
  technique in net-centric environment

31
Q A
32
The End