|
|
|
Javed I. Khan |
|
|
|
ACTIVENET PI Meeting , June 2001, Jackson Hole |
|
|
|
Networking and Media Communications Research Lab |
|
Dept of Math & Computer Science |
|
Kent State University |
|
Sponsor: DARPA/ITO |
|
|
|
|
|
|
|
|
Link Capacity Adaptation |
|
Splices wide-band links with low capacity links. |
|
Application aware rate adaptation. |
|
Greater down-scalability range (4/5th vs.
1/100th) |
|
Perceptual encoding |
|
Interactive data compression with perceptual control over where and how
quality is compromised. |
|
|
|
Node Capacity Adaptation |
|
Allows network junction nodes with unequal CPUs
to perform rate adaptation. |
|
Auto-configuration by motion vector bypass |
|
Reuses motion vector and saves computation
for quality |
|
Cycle sweeping (network spread) |
|
If one node is not enough it spreads the
computation upstream and sweeps additional cycles from active nodes. |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Scalability=deeper knowledge+system symbiosis |
|
Rate adaptation at network level? Only 80%. |
|
We are trying to gain deeper interaction: |
|
Content protocol aware adaptation
(requantization, DCT filtering, etc.) |
|
Content aware adaptation (object detection
based). |
|
Interactive viewer aware adaptation. |
|
Active network insight: |
|
Look at the source of info that forms the base
of adaptation? |
|
video protocol designer |
|
content producer |
|
viewer |
|
above all-- the network resource awareness |
|
Each knowledge component have their distinct
role. How to connecting all
the parties involved? |
|
|
|
|
|
|
|
|
We needed two things: |
|
Collect link and node capacities to find out
optimum transcoding points. |
|
Now set the configuration parameters. |
|
|
|
We consequently designed a network local state
collection and propagation tool |
|
User application sends the propagation and
synthesis math, the harness takes care of the propagation by 2.5 phase
communication. |
|
A building block for transcoder deployment in a
multicast network. |
|
A generalization of SNMP |
|
Instead of P2P it is now graph wide. |
|
Local composition on local MIB2 elements at
active nodes. |
|
|
|
|
|
|
|
|
|
|
|
|
Sits quite at the other end of a narrow link and monitors surfer’s
behavior. Accordingly pre-fetches, caches, transcodes, and performs
customized stream scheduling. |
|
|
|
|
|
|
|
In a large network diversity is ingrained |
|
|
|
Whether it is an adaptive service or not, but
the ingredient components should not make very restrictive assumption about
network resources. |
|
This requires certain behaviors (potential
“Active Network Layer Services) |
|
Network resource awareness |
|
State awareness about the service |
|
Self-organization help (state maintenance
signaling?) |
|
Plug-in in the applications themselves
(footprint). |
|
Seamless reorganization |
|
|
|
|
|
|
|
Performance and Onion Peeling: |
|
Encapsulation does not stop at TCP layer it continues upward… |
|
|
|
[IP[TCP[TS[SEQ[PIC[SLICE[MB]]]]]]] |
|
|
|
Active network have to worry about performance
of access. |
|
To get access to one field-- generally we have
to access. somewhere up to 5-10 other fields! Major performance drainage. |
|
Current protocols = logic + field placement |
|
Proposal = Separate logical definition from locution in packet. |
|
Optimum locution based on access dependency of
the capsules. |
|
|
|
|
|
|
|
Active network feels like very crude bare-bone
architectures. |
|
|
|
Powerful yet simple construction formalism is
critically required to encourage usability of the architecture and
reusability of the efforts spent to develop middleware on them. |
|
|
|
Some of our organized thoughts hinting
solutions: |
|
Language to express ‘deployment constraints’. |
|
Language to express ‘resource requirements’. |
|
Language to express ‘bounds of adaptation’. |
|
“Virtual instrument” |
|
Definitely I am out of time by now! Pl. see our
publications. |
|
|
|
Papers are available at
http://medianet.kent.edu/ |
|
Notes