We enable reliable and dependable self‐adaptations of component connectors in unreliable environments with imperfect monitoring facilities and conflicting user opinions about adaptation policies by developing a framework which comprises: (a) mechanisms for robust model evolution, (b) a method for adaptation reasoning, and (c) tool support that allows an end‐to‐end application of the developed techniques in real‐world domains.

1. A Framework for Robust Control of Uncertainty
in Self-Adaptive Software Connectors
Pooyan Jamshidi
Lero – the Irish Software Engineering Research Centre
School of Computing, Dublin City University
Pooyan.jamshidi@computing.dcu.ie
Supervised by: Dr. Claus Pahl
Environment=D
Environment=D’
Environment=D’
Adapted to satisfy
requirements
while it is runningü Reliable (Robust)
ü Run-time Efficient

2. 2
Direct message passing
(e.g., EJB, CORBA)
Indirect message passing (e.g., JavaBeans, ADLs)
Exogenous connectors (e.g., Reo)

3. 3

4. ⊭ 𝑅
Environment=D
Environment=D’
Environment=D’
⊨ 𝑅
⊨ 𝑅
Adapted to satisfy
requirements
while it is running
ü Reliable
ü Runtime Efficient
4

5. Requirement: R
0 50 100
0
500
1000
1500
0 50 100
100
200
300
400
500
0 50 100
0
1000
2000
0 50 100
0
200
400
600
0 50 100
0
500
1000
0 50 100
0
500
1000
Environment: D
S1
S2
SE
1-X
X
Y
1-Y
Model: S
MonitoringImplementation
Execution
Runtime
Design-time
Reasoning
Self-adaptation
Specification
Specification
S,D⊨R
S,D⊭R
5

6. Quantitative NFRs:
Reliability,
Performance
6
Chapter 4
Chapter 5
Chapter 6

7. 4. Model Calibration
Research Question 1

8. 0.2
0.8
S D T L
S
0 0 10 0
D
6 0 0 0
T
0 6 0 4
L
0 0 2 0
6
CTMC (Continuous-Time Markov Chain)
DTMC (Discrete-Time Markov Chain)
S D T L
S
0 0 1 0
D
1 0 0 0
T
0 0.9 0 0.1
L
0 0 1 0
8

9. PARAM Model Checker
dDdAdDdAdD
dAdA
dAdDdAf
Lost
Lost
Lost
**
*
),,(
2
++
=
8.0
4*424*2
4*4
)4,2,4(
2
=
++
=f
1}"_{":1 <lostmessageRNFR
9

10. 𝑟!,!
#
𝑟!,$
#
𝑟$,!
#
𝑟$,$
#
𝑟!,!
%
𝑟!,$
%
𝑟$,!
%
𝑟$,$
%
Runtime Data
10

11. qEstimate the updated transition matrix (posteriori)
given runtime traces and prior transitions
ØA statistical approach (Bayesian estimator)
ØExtension: assigning weight to the observations
ØRelated work: (Epifani et al., 2009), (Calinescu et al., 2011)
Prior Estimation
@ Design-Time
Posterior Tuning
@ Runtime
Estimate
∑!"#
$
'!(%,'
!
∑!"#
$
'!
Simple Exponential
Smoothing
11

12. NFR1satisfiedNFR1violated
EstimateprobabilityP
Time
p Actual
Estimation
w=1.01
Estimation
w=1.001
Estimation
Standard
Bayes Rule
12

13. 13
Observation (monitoring data) might be noisy
Discrete-time
observations
Only a subset
of the model
Is observed

14. )0(
R xX =
),|( )(
xXRP i
),|( )1(
xRXP i-
))(ˆ,
))(ˆ(
1
( ijij
ii
TN
TR
a
b
+
+
G
)(ˆ k
R
14

15. 0 50 100 150 200 250 300 350 400 450 500
-2
-1
0
1
2
3
4
5
6
7
L1
L2
Norm
SVD
15

16. 0
1
2
3
4
5
6
7
1 2 3 4 5 6 7 8 9 10 11 12
variable Exp. 1 Exp. 2 Exp. 3 Exp. 4 Exp. 5 Exp. 6 Exp. 7 Exp. 8 Exp. 9 Exp. 10 Exp. 11 Exp. 12
𝑺𝒕𝒂𝒕𝒆𝒔 8 8 8 8 8 8 8 8 8 8 8 8
𝑨𝒃𝒔 1 1 1 1 1 1 1 1 1 1 1 1
𝑻 7 7 7 7 7 7 7 7 7 7 7 7
𝚫𝒕 1 1 1 1 1 1 1 1 1 1 1 1
𝑶𝒃𝒔 1 2 3 4 5 10 20 30 40 100 200 500
𝜶 1 1 1 1 1 1 1 1 1 1 1 1
𝜷 1 1 1 1 1 1 1 1 1 1 1 1
𝑴 2000 2000 2000 2000 2000 2000 2000 2000 2000 2000 2000 2000
𝒃 500 500 500 500 500 500 500 500 500 500 500 500
16

17. dA dD
A_Lost
A2B
B2F
F2C
C2F
F2D
dDdAdDdAdD
dAdA
dAdDdAf
Lost
Lost
Lost
**
*
),,(
2
++
=
8.0
4*424*2
4*4
)4,2,4(
2
=
++
=f
1}"_{":1 <lostmessageRNFR
17
<1 √

18. 5. Adaptation Reasoning
Research Question 2

19. 0 0.5 1 1.5 2 2.5 3
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
Region of
definite
satisfaction
Region of
definite
dissatisfactionRegion of
uncertain
satisfaction
Performance Index
Possibility
Performance Index
Possibility
words can mean different
things to different people
Different users often
recommend
different adaptation policies
0 0.5 1 1.5 2 2.5 3
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
Type-2 MF
Type-1 MF
19

20. RobusT2
Initial setting +
adaptation rules +
requirements
environment
monitoring
application
monitoring
Adaptation
actions
Fuzzy Reasoning
Users
Prediction/
Smoothing
20

21. Workload
Response time
0 10 20 30 40 50 60 70 80 90 100
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
x2
uMembershipgrade
0 10 20 30 40 50 60 70 80 90 100
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
uMembershipgrade
=>
=>
mean
sd
21

22. Rule
(𝒍)
Antecedents Consequent
𝒄 𝒂𝒗𝒈
𝒍
Workload
Response-
time
Normal
(-2)
Effort
(-1)
Medium
Effort
(0)
High
Effort
(+1)
Maximum
Effort (+2)
1 Very low Instantaneous 7 2 1 0 0 -1.6
2 Very low Fast 5 4 1 0 0 -1.4
3 Very low Medium 0 2 6 2 0 0
4 Very low Slow 0 0 4 6 0 0.6
5 Very low Very slow 0 0 0 6 4 1.4
6 Low Instantaneous 5 3 2 0 0 -1.3
7 Low Fast 2 7 1 0 0 -1.1
8 Low Medium 0 1 5 3 1 0.4
9 Low Slow 0 0 1 8 1 1
10 Low Very slow 0 0 0 4 6 1.6
11 Medium Instantaneous 6 4 0 0 0 -1.6
12 Medium Fast 2 5 3 0 0 -0.9
13 Medium Medium 0 0 5 4 1 0.6
14 Medium Slow 0 0 1 7 2 1.1
15 Medium Very slow 0 0 1 3 6 1.5
16 High Instantaneous 8 2 0 0 0 -1.8
17 High Fast 4 6 0 0 0 -1.4
18 High Medium 0 1 5 3 1 0.4
19 High Slow 0 0 1 7 2 1.1
20 High Very slow 0 0 0 6 4 1.4
21 Very high Instantaneous 9 1 0 0 0 -1.9
22 Very high Fast 3 6 1 0 0 -1.2
23 Very high Medium 0 1 4 4 1 0.5
24 Very high Slow 0 0 1 8 1 1
25 Very high Very slow 0 0 0 4 6 1.6
Rule
(𝐥)
Antecedents Consequent
𝒄 𝒂𝒗𝒈
𝒍Work
load
Response
-time
M
1
M
2
M
3
M
4
M
5
12 Medium Fast 2 5 3 0 0 -0.9
10 experts’ responses
𝑅-: IF (the workload (𝑥!) is )𝐹.#
, AND the response-
time (𝑥$) is )𝐺.(
), THEN (change the connector mode
to …).
𝑐/01
- =
∑23!
4)
𝑤2
-
×𝐶
∑23!
4)
𝑤2
-
Goal: pre-computations of costly calculations
to make a runtime efficient adaptation
reasoning based on fuzzy inference 22

23. Liang, Q., Mendel, J. M. (2000). Interval type-2 fuzzy
logic systems: theory and design. Fuzzy Systems, IEEE
Transactions on, 8(5), 535-550.
Adaptation Actions
Monitoring Data
23

24. 0 10 20 30 40 50 60 70 80 90 100
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0.5954
0.3797
𝑀
0 10 20 30 40 50 60 70 80 90 100
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0.2212
0.0000
0 10 20 30 40 50 60 70 80 90 100
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
x2
u
0 10 20 30 40 50 60 70 80 90 100
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
u Monitoring data
Workload
Response time
24

25. 0 10 20 30 40 50 60 70 80 90 100
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0.5954
0.3797
0 10 20 30 40 50 60 70 80 90 100
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0.9568
0.9377
25

26. 26

27. 6. Adaptation Execution

28. 𝐶234567689 =< 𝐴 , 𝐵 , 𝐴, 𝑓1, 𝐹𝐼𝐹𝑂, 𝐵
𝐶23456768: =< 𝐴 , 𝐵 , 𝐴, 𝑓1, 𝐹𝐼𝐹𝑂, 𝑁1 , 𝑁1, 𝑓2, 𝐹𝐼𝐹𝑂, 𝐵
𝐶;<=9 =< 𝐴 , 𝑂1, 𝑂2, 𝑂3, 𝐵 ,
{ 𝐴, 𝑠1, 𝑆𝑦𝑛𝑐, 𝑁1 , 𝑁1, 𝑠2, 𝑆𝑦𝑛𝑐, 𝑂1 ,
𝑁1, 𝑃1, 𝐶1, 𝐶𝑇1, 𝑃3, 𝑂3 , 𝑁1, 𝑃1, 𝐶1, 𝐶𝑇1, 𝑃2, 𝑁2 ,
𝑁2, 𝑠3, 𝑆𝑦𝑛𝑐, 𝑂2 , 𝑁2, 𝑠4, 𝑆𝑦𝑛𝑐, 𝐵 } >
28

29. 29

30. 𝐶𝐶5! = {𝐷𝑎𝑡𝑎67, 𝐷𝑎𝑡𝑎829, 𝑀𝑖𝑑𝑑𝑙𝑒𝐿𝑎𝑦𝑒𝑟,
𝑆𝑖𝑚𝑝𝑙𝑒 𝐵𝑢𝑓𝑓𝑒𝑟, 𝑂𝑢𝑡𝑝𝑢𝑡, 𝑂𝑢𝑡𝑝𝑢𝑡1, 𝑂𝑢𝑡𝑝𝑢𝑡2}
𝐹2𝐶 𝐷𝑎𝑡𝑎_𝐼𝑛 =< 𝐴 , 𝐴, 𝑠1, 𝑆𝑦𝑛𝑐, 𝑁1 >
𝐹2𝐶 𝑆𝑖𝑚𝑝𝑙𝑒 𝐵𝑢𝑓𝑓𝑒𝑟 =< 𝑁1, 𝑁2 , 𝑁1, 𝑓1, 𝐹𝐼𝐹𝑂, 𝑁2 >
𝐹/:: = 𝐹9 − 𝐹;
𝐹<=> = 𝐹; − 𝐹9
𝑝/:: = 𝐹2𝐶 𝐹/:: =< 𝑁/::, 𝑅/:: >
𝑝<=> = 𝐹2𝐶 𝐹<=> =< 𝑁<=>, 𝑅<=> >
𝑅9 = 𝑅; + 𝑅/:: − 𝑅<=>
30

31. 31
Structural constructs
Feature-based reasoning
Reconfiguration operations

32. 7. Implementation and
Evaluation
Research Question 3

33. 33

34. 0 50 100
0
500
1000
1500
0 50 100
100
200
300
400
500
0 50 100
0
1000
2000
0 50 100
0
200
400
600
0 50 100
0
500
1000
0 50 100
0
500
1000
34

35. 35

36. 36

37. 0
1
2
3
4
5
x 10
-3
1 2 3 4 5 6 7 8
0
0.05
0.1
0.15
0.2
0.25
1 2 3 4 5
37

38. 0.05
0.1
0.15
0.2
0.25
0.3
0.35
Type-1 FLS Type-2 FLS
RMSE
• The rule reduction reduced the rules
quite considerably.
• IT2 FLCs are more robust due to less
mean error and less variation in the
estimation error.
• T1 FLCs in some realization drop more
rules in comparison with the IT2 FLCs.
• IT2 FLC original designs can be designed
with less rules.
38

39. 0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
1% 2% 3% 4% 5% 6% 7% 8% 9% 10%
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1% 2% 3% 4% 5% 6% 7% 8% 9% 10%
0
0.2
0.4
0.6
0.8
1
1% 2% 3% 4% 5% 6% 7% 8% 9% 10%
Noise Amplitude (α=0.5)
RMSE
α=0.95α=0.7
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
1% 2% 3% 4% 5% 6% 7% 8% 9% 10%
T1 FLCIT2 FLC
39

40. 8. Conclusions

41. 41

42. 42

43. 43

44. ⊭ 𝑅
Environment=D
Environment=D’
Environment=D’
⊨ 𝑅
⊨ 𝑅
Adapted to satisfy
requirements
while it is running
ü Reliable
ü Run-time Efficient
dA dD
A_Lost
A2B
B2F
F2C
C2F
F2D
dDdAdDdAdD
dAdA
dAdDdAf
Lost
Lost
Lost
**
*
),,(
2
++
=
8.0
4*424*2
4*4
)4,2,4(
2
=
++
=f
1}"_{":1 <lostmessageRNFR
RobusT2 Initial setting +
adaptation rules +
response-time SLA
environment
monitoring
application
monitoring
Adaptation
actions
Fuzzy Reasoning
Users
Prediction/
Smoothing
44
Thank you!

46. 46
Evolved
ArchitectureC1 C2 Add Remove
Modify
C1 C2
CMSource
Architecture
Change description
Change verification
Change accommodation

47. RQ0: “How to enable a robust and runtime efficient self-
adaptation for software connectors and make them reliable
to be used in Open environments?”
R
~
Knowledge
Specification
Uncertainty
Measurement
Inaccuracy
Naeem Esfahani and Sam Malek,
“Uncertainty in Self-Adaptive
Software Systems”
47

48. 48Kung-Kiu Lau, et al. “Exogenous Connectors for Software Components”

49. dA: 4 dD: 2
A_Lost: 4
dA dD
A_Lost
A2B
B2F
F2C
C2F
F2D
Variable model parameters
Fixed model parameter
Rate of
message output
Rate of
message lost
Rate of
message input
49

50. 0.84
0.031
0.056
50C. Ghezzi, V. PanzicaLa Manna, Alfredo Motta, G. Tamburrelli, "QoS Driven Dynamic Binding
in-the-many", QoSA 2010, Prague, June 23-25, 2010.

51. 51

52. 52

53. 53
NFR1
satisfied
NFR1
violated

54. Reuse
2
54

55. 55

56. 56
Observation
Density
weighted samples from previous time step
Sampling
Approximated by
Importance Sampling

57. Posterior Distribution Correlogram Sample Path
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
0
100
200
300
400
500
600
700
800
900
1000
0 200 400 600 800 1000
-0.2
0
0.2
0.4
0.6
0.8
Lag
SampleAutocorrelation
Sample Autocorrelation Function
0 200 400 600 800 1000
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
Heavily skewed:
mode vs. mean
No alarming pathology
in the sampling
Quick convergence to
stationary distribution
57

58. 58

59. 59

60. 0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
5
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
5
1.5
2
2.5
3
3.5
4
output1
60

61. 61

62. 0 50 100
0
500
1000
1500
0 50 100
100
200
300
400
500
0 50 100
0
1000
2000
0 50 100
0
200
400
600
0 50 100
0
500
1000
0 50 100
0
500
1000
62

63. 0 10 20 30 40 50 60 70 80 90 100
-500
0
500
1000
1500
2000
Time (seconds)
Numberofhits
Original data
betta=0.10, gamma=0.94, rmse=308.1565, rrse=0.79703
betta=0.27, gamma=0.94, rmse=209.7852, rrse=0.54504
betta=0.80, gamma=0.94, rmse=272.6285, rrse=0.70858
0
0.2
0.4
0.6
0.8
1
1.2
1.4
Big spike Dual phase Large variations Quickly varying Slowly varying Steep tri phase
0 50 100
0
500
1000
1500
0 50 100
100
200
300
400
500
0 50 100
0
1000
2000
0 50 100
0
200
400
600
0 50 100
0
500
1000
0 50 100
0
500
1000
RootRelativeSquaredError
63

64. SUT Criteria Big spike Dual phase
Large
variations
Quickly
varying
Slowly
varying
Steep tri
phase
RobusT2Scale
𝑟𝑡?@% 973ms 537ms 509ms 451ms 423ms 498ms
𝑣𝑚 3.2 3.8 5.1 5.3 3.7 3.9
Overprovisioning
𝑟𝑡?@% 354ms 411ms 395ms 446ms 371ms 491ms
𝑣𝑚 6 6 6 6 6 6
Under
provisioning
𝑟𝑡?@% 1465ms 1832ms 1789ms 1594ms 1898ms 2194ms
𝑣𝑚 2 2 2 2 2 2
SLA: 𝒓𝒕 𝟗𝟓 ≤ 𝟔𝟎𝟎𝒎𝒔
For every 10s control interval
•RobusT2 is superior to under-provisioning in terms of
guaranteeing the SLA and does not require excessive
resources
•RobusT2 is superior to over-provisioning in terms of
guaranteeing required resources while guaranteeing the SLA 64

65. 65

66. 66

67. 67
Rebinding between different providers
Increase/Decrease of channel size
Change of channel type

68. 3. Positioning

69. Framework
Source of Uncertainty Feedback Control Loop (MAPE-K)
EvaluationAdaptation
policy
Noisy
data
Simplificatio
n
Change
enactment
Users in
the loop
Dynamic
environment
M A P E K
RequirementSpecification
RELAX Fuzzy goal model Case study
AutoRELAX Fuzzy goal model
Experimental
study
FLAGS Fuzzy goal model Example
Goal-Driven Self-Optimization Prob. Prob. Prob. goal reasoning goal model
Experimental
study
REAssuRE Fuzzy goal reasoning goal model Example
(N Bencomo & Belaggoun, 2013) Prob. goal reasoning
Decision
model
Experimental
study
C/E/I
RCU (This Work) Fuzzy Prob. Control √ (Bayesian learning)
constraint
evaluation
Fuzzy reasoning
Mode
change
Markov
models +
Fuzzy rule
Experimental
study
Partial satisfaction of requirements @ design-time
Resolution @ runtime
Claim
Goal realization
69

70. Internal
Rainbow Prob. Prob. √
constraint
evaluation
√
Architecture
model
Experimental
study
POISED Fuzzy Fuzzy Fuzzy optimization
Architecture
model
Experimental
study
(Cámara et al., 2014) Prob. game analysis
Architecture
model
Experimental
study
ADC Prob. utility reasoning Utility Case study
Framework
Source of Uncertainty Feedback Control Loop (MAPE-K)
EvaluationAdaptation
policy
Noisy
data
Simplificatio
n
Change
enactment
Users in
the loop
Dynamic
environment
M A P E K
C/E/I
RCU (This Work) Fuzzy Prob. Control √ (Bayesian learning)
constraint
evaluation
Fuzzy reasoning
Mode
change
Markov
models +
Fuzzy rule
Experimental
study
Use different theories and reasoning mechanisms to
determine the impact of system change on the quality
properties…e.g., replacing a component on response-
time
70

71. External
FUSION Prob. Prob.
√
(learning)
√
Feature
model
Experimental
study
RESIST Prob. Prob. Prob.
√
(learning)
√
Markov
models
Experimental
study
ADAM Prob.
√
(learning)
√
Markov
models
Experimental
study
KAMI Prob.
√
(learning)
constraint
evaluation
Markov
models
Experimental
study
Veritas/Loki Prob.
test case
verification
test plan
verification;
optimization
Test cases
Experimental
study
C/E/I
RCU (This Work) Fuzzy Prob. Control √ (Bayesian learning)
constraint
evaluation
Fuzzy reasoning
Mode
change
Markov
models +
Fuzzy rule
Experimental
study
Framework
Source of Uncertainty Feedback Control Loop (MAPE-K)
EvaluationAdaptation
policy
Noisy
data
Simplificatio
n
Change
enactment
Users in
the loop
Dynamic
environment
M A P E K
White-box, black-box or gray-box learning approaches
to mitigate environmental uncertainty
71

72. Control
(Antonio Filieri et al., 2014) Control
controller
synthesis
Regression
models
Experimental
study
(Zhu et al., 2009) Control
integral
controller
Regression
models
Experimental
study
C/E/I
RCU (This Work) Fuzzy Prob. Control √ (Bayesian learning)
constraint
evaluation
Fuzzy reasoning
Mode
change
Markov
models +
Fuzzy rule
Experimental
study
Framework
Source of Uncertainty Feedback Control Loop (MAPE-K)
EvaluationAdaptation
policy
Noisy
data
Simplificatio
n
Change
enactment
Users in
the loop
Dynamic
environment
M A P E K
Fuzzy control (knowledge-based) vs. classic (model-based)
Increasing attention in SE community, Dagstuhl seminar,
ICSE’14,…
72

73. Design-time
GuideArch Fuzzy (utility) --
Optimization
(arch. selection)
-- Case study
EAGLE Prob. -- Goal verification Synthesis -- Example
MAVO --
Partial model
reasoning
-- Case study
(H. Yang et al., 2012) --
Machine
learning
Rule reasoning --
Experimental
study
(Arora et al., 2012) --
Feature
interaction
-- Case study
(Letier & van Lamsweerde, 2004) Prob. --
Partial goal
verification
--
(Letier et al., 2014) --
Monte-Carlo
simulation
Pareto-based
optimization
--
Experimental
study
Framework
Source of Uncertainty Feedback Control Loop (MAPE-K)
EvaluationAdaptation
policy
Noisy
data
Simplificatio
n
Change
enactment
Users in
the loop
Dynamic
environment
M A P E K
C/E/I
RCU (This Work) Fuzzy Prob. Control √ (Bayesian learning)
constraint
evaluation
Fuzzy reasoning
Mode
change
Markov
models +
Fuzzy rule
Experimental
study
User involvement and optimization based approaches
may not be necessarily applicable for runtime reasoning
73

74. 74

75. 75

76. 76
Competing consumers Prioritized requests
Pipes and filters Load leveling

77. 77
Request scheduling
Multi-cloud integration
Hybrid integration

78. 78

79. 79
1
~
U
2
~
U
3
~
U

80. 80

81. 81
Design science

82. 82

83. 83

84. 84

85. 85

86. 86

87. 87