Synchronization inspired by fireflies

advertisement
University of Zagreb, Croatia
Synchronization inspired by fireflies
Iva Bojić
University of Zagreb, Croatia
Faculty of Electrical Engineering and Computing
Department of Telecommunications
Summer School of Science 2012
August 7, 2012, Višnjan, Croatia
Round of applause
use your hands and move your body 
University of Zagreb, Croatia
S3++ 2012
13 April 2015
1 of 19
Group of people
same rhythmus
University of Zagreb, Croatia
S3++ 2012
13 April 2015
2 of 19
Heterogeneous Machine-to-Machine Systems
same time
University of Zagreb, Croatia
S3++ 2012
13 April 2015
3 of 19
Logical questions
can we go home? 
University of Zagreb, Croatia
1.
How hard it is to synchronize different clocks?
2.
Why do we need time synchronization?
3.
How can we achieve time synchronization?
S3++ 2012
13 April 2015
4 of 19
Outline
try to give logical answers…
University of Zagreb, Croatia
Research motivation
Biologically-inspired computing
Firefly-inspired synchronization
Results from the laboratory setting
Conclusion
S3++ 2012
13 April 2015
5 of 19
Computer clock
how does it work?
University of Zagreb, Croatia

How is this possible?
Quartz crystals are manufactured
for frequencies from a few tens
of kHz to tens of MHz
A clock is an electronic device that
counts oscillations in a crystal at a
particular frequency
S3++ 2012
13 April 2015
6 of 19
Problem
no global notion of time
University of Zagreb, Croatia
 In distributed systems each
node has its own clock and
its own notion of time
 In practice these clocks
drift apart accumulating
errors over time (1 second
every 11 days)
S3++ 2012
13 April 2015
7 of 19
Need for (speed?)
time
University of Zagreb, Croatia
 Global notion of time is prerequisite for:
 common
resource sharing (e.g. channel)
 depend events tracking (e.g. consistency
of distributed
databases)
 simultaneous events detection (e.g. data collection)
Frequency division multiple access
S3++ 2012
Time division multiple access
13 April 2015
8 of 19
Time synchronization
different algorithms
University of Zagreb, Croatia
 Time
synchronization provides a common time scale for
local clocks of nodes in distributed systems
B. Sundararaman, U. Buy and A. D. Kshemkalyani: Clock synchronization for wireless sensor networks: A Survey, Ad Hoc Networks 3, pp. 281-323 (2005)
S3++ 2012
13 April 2015
9 of 19
Biologically-inspired computing
biology applied in distributed systems
University of Zagreb, Croatia
 Nature is a enormous and a highly
complex system
 processes
are done without any centralized control
 processes
are self-sustainable and self-organized
 Self-organization
is a process where some form of global
order arises out of the local interactions between the
components of an initially disordered system
S3++ 2012
13 April 2015
10 of 19
Self-synchronization
in nature
University of Zagreb, Croatia
a) fireflies
S3++ 2012
b) neurons
13 April 2015
c) heart cells
11 of 19
Self-synchronization
in humans
University of Zagreb, Croatia
S3++ 2012
13 April 2015
12 of 19
Pulse coupled oscillators model
one firefly
University of Zagreb, Croatia


If oscillators are not coupled, their state variables change following
only their own excitations
xi denotes state variable
xi
1
flash
 ti* denotes a moment
i-th oscillator flashes
ex
ci t
at
io
n
xi(t) = fi(t)
flash
threshold
when
0
ti*= T
2T
t
R. E. Mirollo and S. H. Strogatz. Synchronization of pulse-coupled biological oscillators. SIAM J. Appl. Math. 50: pp.1645-1662 (1990)
S3++ 2012
13 April 2015
13 of 19
Pulse coupled oscillators model
two fireflies
University of Zagreb, Croatia

If oscillators are coupled

xi
state variable xi is adjusted upon the
1
flash
threshold
εij
ta
tio
ci
t*i T
flash
flash
threshold
0
xj
ϵij is a coupling constant
flash
n
1
t
2T
gij(t) is a coupling function between
i-th and j-th oscillators
ex

ci
ta
tio

xi(t) = fi(t) + ϵij gij(t)
ex

εij
n
reception of flashes from the others
flash
εji
0
t*j
T
εji
t
2T
R. E. Mirollo and S. H. Strogatz. Synchronization of pulse-coupled biological oscillators. SIAM J. Appl. Math. 50: pp.1645-1662 (1990)
S3++ 2012
13 April 2015
14 of 19
Pulse coupled oscillators model
limitations
University of Zagreb, Croatia
 Pulse coupled oscillators
model assumptions
 no
oscillators with a faulty behavior that desynchronizes the
network
 oscillators are connected in a fully-connected
network
 oscillators
cannot join or leave the network nor change their
positions in the network (i.e. no mobility)
 oscillators are the same
 no delays
S3++ 2012
(i.e. have same frequencies)
in the message exchange among oscillators
13 April 2015
15 of 19
Robustness
with oscillators with faulty behavior
University of Zagreb, Croatia


We embedded a cryptographic mechanism in the pulse coupled
oscillators model to ensure robustness
We used the logical operation exclusive disjunction (i.e. XOR)

provides protection from an attack

does not have a negative effect on the time needed for synchronization
S3++ 2012
13 April 2015
16 of 19
Results
robustness
University of Zagreb, Croatia
S3++ 2012
13 April 2015
17 of 19
Conclusions
can we go home?  NOW WE CAN!!!!!!!!!!!!!!!!!!!
University of Zagreb, Croatia
1.
How hard it is to synchronize different clocks?
2.
Why do we need time synchronization?
3.
How can we achieve time synchronization?
S3++ 2012
13 April 2015
18 of 19
Questions?
University of Zagreb, Croatia
S3++ 2012
13 April 2015
19 of 19
Summation
exclusive disjunction
University of Zagreb, Croatia
Input
A
B
0
0
0
1
1
0
1
1
S3++ 2012
Output
0
1
1
0
13 April 2015
20 of 19
Download