Transcript Présentation PowerPoint

Laboratoire LSR
Logiciels Systèmes Réseaux
Proactive Services
Nguyen Hoa Binh
LSR-IMAG
Grenoble
Outline
 Proactive Services
 Active Node Architecture
 Applications
 Implementation
 Future Works
Why Proactive Services ?
service request
active
network
service state
client
server
packet flow
active node
 End-point Applications need active services from networks
 Active services need information about states from end-point
application
 Active services need information about states from networks
 User role needs to be reduced to supervision
What is a proactive service?
server state
active node
state
client state
Proactive
service
active node
network state
network state
packet flow in the network
 Active service able to react to the changes in the environment
without the intervention of the user
 Get users out from interaction loop – react faster than humain
speed
Proactive Service Behavior
active node
active node
Proactive
service
Proactive
service
uninstall
filter
state
change
packets
install
filter
A
active node
Proactive
service
Proactive
service
packets
active node
packets
B
state
change
C
install
filter
packets
state
change
D
A – when a service is activated, the state of the environment is such
that no processing is required. The service sleeps and packets are
not passed to the service
B – an event signaling a congestion link is sent to the service. It
wakes up and installs an appropriate packet filter. Packets are
passed to the service
C- when the state changes again (congestion disappears), another
event is sent to the service that uninstalls the filter so that packets are
no longer intercepted. The service returns to sleep.
D - the whole process may repeat
Active Node Architecture
Execution/
Adaptation
Management
user
Control
interaction
subscribe
events
Proactive service
events
Monitor
services:
user space
probe
filter
install
packets in
Monitoring/
Notification
probe
filter
Matching Engine
packets out
 Association of proactive services with chosen packet flows is done
dynamically by the services themselves
 Monitors are able to detect varying conditions in the environment
(network, active nodes, services, users)
 An asynchronous notification mechanism allows monitors to notify
proactive services about events
forwarding:
kernel space
Example of Caching Service
actived but
sleeping
mpeg1 video
video server
wireless link
probe
packets
wireless
client
active router
PAUSE
video server
caching
service
presence
monitor
disconnect
event
wireless
client
active router
PLAY
video server
presence
caching monitor
service
caching presence
service monitor
event
active router
wireless link
probe
packets
wireless
client
Session Migration Service
Directory Server
Content
request
response
network state
data transmission
Client
 What
happens when the connection between the
client and the server is degraded ? Server
overloaded ? Server disconnected ? etc…
Server
Session Migration Service
Directory Server
server state
client state
network
proactive
service
state
data transmission
server state
Server 1
Client
Server 2
 Proactive
service establishes new data session
with other servers
Services for Ubiquitous Computing
Video Server
MPEG-H263
Transcoder Service
Large Screen
Room
MPEG
Services for Ubiquitous Computing
Video Server
MPEG-H263
Transcoder Service
H263
Large Screen
Room
MPEG
Services for Ubiquitous Computing
Video Server
MPEG-H263
Proactive Service
Event
H263
Large Screen
Room
Event
MPEG
Services for Ubiquitous Computing
Video Server
MPEG-H263
Proactive Service
MPEG
H263
Event
Large Screen
Room
Implementation
user
process
user
process
user
process
libipq
iptables
user space
ip_queue
rules
packets
in
2
5
Netfilter
packets out
1
routing
3
4
kernel
Routeur running Linux kernel 2.4
Implementation
 Netfilter and ip_queue are modified to support forwarding
of different packet flows to different user space processes
 Proactive services run as user space processes and can
be written using C or Java language.
 An example of a user space active network node
 Node monitors the resources used by proactive services
Future Works
 Provide a high level language for specifying proactive services
reason in terms of monitors, events, and actions
 Compiled into an implementation language such as C or Java
 Experiment with proactive services for different applications
 ubiquitous computing
 content distribution
 sensor networks
Dynamic Routing Service
Telephone
Networks