Transcript 14.30 h, 10 minute presentation per PhD project (by the supervisors)

Kickoff meeting
15 September 2015
Eindhoven
Agenda kick-off meeting Big Software on the Run, September 15
14.00 h, welcome, all participants introduce themselves briefly;
14.20 h, overview project (Wil)
14.30 h, 10 minute presentation per PhD project (by the supervisors)
15.30 h, data
16.00 h, coffee/tea break
16.30 h, collaboration
a.
industry contacts
b.
other researcher
16.45 h, budget
a.
infrastructure
b.
workshops/symposia
c.
summer/winter schools
17.30 h, planning
a.
next meeting BSoR, total group
b.
follow up meeting tracks
c.
vacancies
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
Aalst, W.M.P. van der
Wijk, J.J. van
Dongen, B.F. van
Wetering, H.M.M. van de
Liu, C.
bart.postma@inria .fr
R. [email protected]
Arie van Deursen ([email protected])
Sicco Verwer ([email protected])
Zekeriya Erkin - EWI ([email protected])
'[email protected]'
[email protected]
'[email protected]'
m.i.a.stoelinga@ utwente.nl
[email protected]
Leemans, M.
Caroline de Wit ([email protected])
Tamara Brusik ([email protected])
[email protected]
[email protected]
©Wil van der Aalst & TU/e (use only with permission & acknowledgements)
©Wil van der Aalst & TU/e (use only with permission & acknowledgements)
growing scale
growing complexity
continuous evolution
of software
continuously changing
environment
increasing diversity
threats to security and
thrust
not-standalone
interconnected
distributed
cloud
Hadoop
sensor
late composition
updates
apps
growing scale
growing complexity
continuous evolution
of software
continuously changing
environment
increasing diversity
multiple platforms
many versions
many configurations
hardware changes
OS, network, etc.
unanticipated use
threats to security and
thrust
we rely on
software in a
hostile open world
BSR: Big Software on the Run
RECOMMENDATION
T4
DIAGNOSTICS
SOFTWARE DEVELOPMENT
DISCOVERY
BIG SOFTWARE
T1
EVENT DATA
x
RUNNING SYSTEMS
VISUALIZATIONS
CONFORMANCE
CHECK
KNOWLEDGE
!
T2
PREDICTION
PROCESS MODELS
T3
EXTRACTION
BIG INFRASTRUCTURE
T5
BSR: Big Software on the Run
Software in its natural habitat!
€ 34,722,000
€ 1,999,989
Wil van der Aalst &
Jack van Wijk
Arie van Deursen &
Inald Lagendijk
Jaco van de Pol &
Marieke Huisman
Process Mining
Eindhoven
University of
Technology
(TU/e)
Visual Analytics
Process Mining
Visual Analytics
Automatically
Discovering Behavioral
Software Models from
Software Event Data
Model-based
Visualization of
Software Event Data
Parallel Checking and
Prediction
Exceptional Patterns
3TU.BSR
Software Engineering
Concurrent
Verification and
Validation
Software Engineering
Delft University
of Technology
(TUD)
University of
Twente
(UT)
Verification and
Privacy and Security
6 PhDs + 3 postdocs
2015-2019
Privacy Preserving Online Conformance
Checking
Privacy and Security
Validation
Concurrent
Software
Monitoring Concurrent
Software
Software
Eindhoven
University of
Technology
(TU/e)
Process Mining
Cong Liu
Visual Analytics
Bart Postma
Process Mining
Visual Analytics
Automatically
Discovering Behavioral
Software Models from
Software Event Data
Model-based
Visualization of
Software Event Data
Vincent Bloemen
Parallel Checking and
Prediction
Exceptional Patterns
3TU.BSR
Software Engineering
Concurrent
Verification and
Validation
Software Engineering
Delft University
of Technology
(TUD)
Software
University of
Twente
(UT)
Gamze Tillem
Verification and
Privacy and Security
Privacy Preserving Online Conformance
Checking
Privacy and Security
Validation
Concurrent
Software
Monitoring Concurrent
Software
Arnd Hartmanns (UT)
Process Mining
Visual Analytics
Software Engineering
Automatically
Discovering Behavioral
Software Models from
Software Event Data
(part of Track T1)
Van der Aalst & Van
Deursen
1
C
Model-based
Visualization of Software
Event Data
(part of Track T1)
Van Wijk & Huisman
Exceptional Patterns
(part of Tracks T1 and T4)
Van Deursen & Van Wijk
C
C
Monitoring Concurrent
Software
(part of Tracks T1 and T2)
Huisman & Lagendijk
C
Privacy Preserving Online Conformance
Checking
(part of Track T2)
Lagendijk & Van de Pol
C
Concurrent Software
2
Verification and
Validation
C
1
C
C
2
2
1
2
C
1
1
2
C
Parallel Checking and
Prediction
(part of Tracks T2 and T3)
Van de Pol & Van der
Aalst
2
C
C
C
Privacy and Security
C
1
Automatically Discovering Behavioral Software Models from Software Event Data (part of Track T1) Van der Aalst & Van Deursen
Process models and user interface workflows underlie the functional specification of almost every substantial software system.
However, these are often left implicit or are not kept consistent with the actual software development. When the system is utilized,
user interaction with the system can be recorded in event logs. After applying process mining methods to logs, we can derive
process and user interface workflow models. These models provide insights regarding the real usage of the software and can
enable usability improvements and software redesign. In this project, we aim to develop process discovery techniques specific for
software. How can domain knowledge and software structure be exploited while mining? How to discover software patterns and
anti-patterns?
Model-based Visualization of Software Event Data (part of Track T1) Van Wijk & Huisman
Visualization can be a powerful means for understanding large and complex data sets, such as the huge event streams produced
by running software systems. During explorative analysis (T1) experts have to be enabled to see what patterns occur, during
monitoring (T2) anomalous events and patterns have to be detected, where in both cases we can exploit the unique capabilities of
the human visual system. However, simply showing events as a sequence of items will fall short because of lack of scalability. The
challenge is to enable users to specify what they are interested in, and to show only a limited subset of the data, using filtering,
aggregation, and abstraction. We propose to enable users to define models for this, ranging from simple range filters to process
models. We will study which (combinations of) models are most appropriate here, such that occurrences of events, temporal and
logical patterns, and the relations between occurrences and attributes of events can be detected, and to facilitate analysts to
define and check hypotheses on patterns.
Exceptional Patterns (part of Tracks T1 and T4) Van Deursen & Van Wijk
A particularly challenging phenomenon in software development are 'exceptions'. Most programming is focused on 'good weather
behavior', in which the system works under normal circumstances. Actual deployment however, often takes place in a changing or
unexpected environment. This may lead to exceptions being raised by the application, which should be handled by the application.
Unfortunately, predicting such exceptional circumstances is often impossible. Consequently, developers have difficulty
adequately handling such exceptions. Some exceptions are simply swallowed by the applications, others are properly logged, and
yet other may lead to unpredictable behavior. To resolve this, we propose to analyze log files for 'exceptional patterns' -- patterns
that hint at the presence of exceptions. To find such patterns, we propose to use visualization techniques applied to log data and
stack traces. Furthermore, we will investigate ways to predict future occurrences of exceptions, and recommendations on how to
improve exception handling in the code base.
Monitoring Concurrent Software (part of Tracks T1 and T2) Huisman & Lagendijk
The goal is to develop a monitoring system for concurrent software. Making monitoring transparent is the big challenge:
monitoring should not affect program behavior. A general purpose approach will be designed, based on local annotations and
global properties. Runtime monitoring is essential to check conformance of concurrent software during deployment (T2). At the
same time, runtime monitoring provides insight in low-level software events, generating a continuous data stream of events that
feeds discovery (T1). With process mining and visualization technology in Eindhoven, we will explore the scope of concurrent
software monitoring.
Privacy Preserving On-line Conformance Checking (part of Track T2) Lagendijk & Van de Pol
Privacy enhancing techniques have been applied dominantly to data analysis problems (such as pattern recognition) and
multimedia algorithms (such as recommendation engines). The goal of privacy preserving on-line conformance checking is to
research the problem of privacy and security protection in software engineering for the first time. The central problem is that
conformance checking algorithms may need to operate on event data that is sensitive in some way, for instance, contains userrelated information. Such data can be anonymized or encrypted for protection, yet this might affect the accuracy of the
conformance checking procedure. It will therefore be necessary to find an acceptable trade-off between the level of protection, the
utility of the results obtained from the privacy-enhanced version of the conformance checking algorithm, and the additional
computational overhead introduced by the anonymization or encryption process.
Parallel Checking and Prediction (part of Tracks T2 and T3) Van de Pol & Van der Aalst
Based on the models discovered by online observations (Track 1), the goal of this research project is to develop scalable
technology for predicting future system behavior (Track 3). Assuming that the system’s components will behave similar to the
process models learnt so far, (quantitative) model checking techniques will be applied to explore possible runs and interactions of
the integrated system. In order to support online recommendations (Track 4), the model checking results should be available
nearly instantaneously. This calls for parallel, scalable algorithms that will be run on BSR.cloud infrastructure in Twente. Large
scale distributed experiments will be run on DAS5.
Begroting BSoR 2014-2018
2014
Beste Eveline,
2015
2016
2017
2018
2019
Totaal
94,293 €
70,311 €
94,293 €
35,156
47,147 €
€
377,172
210,934
TU/e
2 Promovendi
1 Postdoc
€0 €
€0 €
47,147 €
35,156 €
94,293 €
70,311 €
Computer
Conferences, travel
Other
€0 €
€0 €
€
4,000
2,500 €
4,250 €
5,500 €
4,250 €
5,500
4,250
€
€
€
4,000
13,500
12,750
€0 €
20,000 €
60,000 €
20,000
€
100,000
€
€
50,000
768,356
45,222 €
€
361,776
214,568
€
€
€
€
45,222 €
4,000
20,000
16,350
616,694
87,444 €
35,806
43,722 €
€
349,776
214,838
€
€
€
43,722 €
4,000
27,500
18,825
614,939
Infrastructure voor gehele project,
inclusief software, speciale
hardware, website, …
Organisatiekosten workshops,
symposia, guests, …
TU/e totaal
€0 €
€0 €
10,000 €
20,000 €
123,053 € 254,354 €
20,000
214,354 €
129,449
€
47,147
UT
2 Promovendi
1 Postdoc
€0 €
€0 €
45,222 €
35,761 €
90,444 €
71,523 €
90,444 €
71,523 €
Computer
Conferences, travel
Other
UT totaal
€0 €
€0 €
€
€0 €
4,000
4,000 €
8,000 €
5,450 €
5,450 €
94,433 € 175,417 €
8,000
5,450
175,417 €
TUD
2 Promovendi
1 Postdoc
€0 €
€0 €
43,722 €
35,806 €
87,444 €
71,613 €
87,444 €
71,613 €
Computer
Conferences, travel
Other
TUD totaal
€0 €
€0 €
€
€0 €
4,000
5,500 €
11,000 €
6,275 €
6,275 €
95,303 € 176,332 €
11,000
6,275
176,332 €
3TU TOTAAL
€ 0 € 312,789 € 606,103 € 566,103 € 378,904 € 136,091 € 1,999,989
90,444 €
35,761
126,205
123,250
€
Zoals vandaag telefonisch besproken.
1. In de word file stuur ik wat tekst (ook met dank
aan Ine) voor het rapport (zie bijlage).
2. We gaan er vanuit dat (a) de onderstaande
begroting leidend is, (b) dat wel geschoven mag
worden met de aanstellingen (maar voor eind
2019 moeten aanstellingen afgelopen zijn), (c)
de andere kosten moeten voor eind 2017
gemaakt zijn (de verdeling over de jaren is
richtinggevend maar zeker niet hard, PhDs gaan
dit jaar bijvoorbeeld waarschijnlijk niet naar
conferenties en er zijn geen workshops,
summerschools, etc.).
3. Je probeert dit formeel af te stemmen met
IJsbrand Haagsma <[email protected]>.
Het is essentieel dat voor de kick-off meeting,
“the rules of the game” helder zijn !!
4. De verdeling is zoals deze in onderstaande
begroting staat. Universiteiten die een
reserveringsbeleid volgen kunnen proberen
reiskosten etc. voor 2018 en 2019 eerder te
maken. Actiepunt voor elke universiteit
afzonderlijk dit te implementeren indien
mogelijk. Voor TU/e gaat Jacqueline den Braven
dit onderzoeken. Ook afstemmen met Suwita
van den Biggelaar ([email protected])
5. NIRICT gelden voor SRA Data Science zijn
toegekend voor tweede helft van het jaar. Hier
volgt nog een e-mail over.
Groeten, Wil.