A Weakly Coupled Adaptive Gossip Protocol for Application Level - - PowerPoint PPT Presentation

1

A Weakly Coupled Adaptive Gossip Protocol for Application Level Active Networks

Ibiso Wokoma, Ioannis Liabotis, Ognjen Prnjat, Lionel Sacks, Ian Marshall Department of Electronic Engineering, University College London

In association with

2

Problem Space

Policy management of distributed networks requires

that system and node policies are updated regularly

The distribution of the updates needs to be

asynchronous scalable, reliable and independent of the underlying

topology

implemented without centralized planning

The proposed protocol satisfies these requirements

3

Gossip protocols

Efficient method of distributed information

dissemination in wide-area networks

Delayed propagation that provides weak

consistency and better availability

e.g. Time Stamped Anti-Entropy (TSAE)

Log B Sum B

A B C

3 2 5 8 5 14 7

A B C

Log A, B Sum A, B

3 5 6 2 8 7 5 10 14 7

After anti-entropy

Node A Node B 14 7

Log A

A B C

Sum A

3 6 2 8 10 4 5

Before anti-entropy

4

Firefly synchronization

5

Firefly synchronization

timer = 1 timer = 15 timer = 10 timer = 7 t = 1 epoch timer = 4 timer = 0 timer = 14 timer = 10 t = 3 epochs timer = 2 timer = 16 timer = 11 timer = 8 t = 2 epochs

A biologically inspired model based on

fireflies was designed

Fireflies alter their internal timers to flash at

the same rate as their neighbours

An example of self organization using a non-

deterministic distributed mechanism

6

Firefly/Gossip Model (1)

The functional requirements for this model are:

low network traffic and synchronization time scalable efficiency in different networks

To achieve this, each node is given a:

timer flash interval – period the node waits before it

“flashes” i.e. communicates with its neighbours

event record – a list of the node’s neighbours with an

indication of how frequently they have flashed

hash table – a list of the policies added to the node

with a timestamp for each

7

Firefly/Gossip Model (2)

Randomize timers For each epoch and for each node in the network, decrement timers The node broadcasts its’ hash table to its neighbours who add the flashing node to their event record if its not already there and compare hash tables Is timer larger than 0? Has a new policy been found? Decrease the flash interval Increase the flash interval Reset timer to the value of the flash interval Subsequent flashes result in changes to the event records; if the neighbours have not flashed much in comparison to the node being analysed then it is removed from the list yes no yes no

8

Optimization tests

9

Scalability tests

10

Topology tests

11

Conclusions

An autonomous and self-organized method of

distributing updates in systems such as active networks, GRID and sensor networks

The mechanism is robust in response to node

failures and performs better than TSAE

Further work could involve assessing how the

algorithm works in small worlds and scale- free topologies and with different policy arrival rates