This document discusses different model representations for large meta-model based datasets. It compares object-by-object representation to fragmentation strategies. Fragmentation breaks models into multiple fragments stored separately. The document evaluates different fragmentation strategies through theoretical analysis and implementation tests. It also compares part-of-source and relational representations and discusses applications of model fragmentation including software engineering and scientific data analysis.

1. Markus Scheidgen
Model representations for
large meta-model based data-sets
￭ Introduction: Technological spaces and model representations
￭ Comparison of representation
￭ Implementation
￭ Application
1

2. Introduction:
Technological Spaces
2
Software Models
Code
reverse engineering
code generation
XML
persistence / exchange
databases
persistence/versioning
processing
(via ORMs: e.g. JPA)
Objects
(e.g. POJOs)
debugging/profiling
reflection
runtimemodeling
processing (e.g. dom/jaxb)
exchange (e.g. in web-services) xslt/xsl/
xquery/xpath
model-transformation/
-constraints/-queries
static analysis/compilation/
refactoring
SQL
running programs
other data
otherdata
otherdata
otherda
ta
ot
herdata

3. Introduction:
State of the Art
3
Meta-Models
Models
Schemas
XML
Gammars
Code
Classes
Objects
ER-Schemas
Relational Data
*
visualization and editing
by human users
processing in computer programs
exchange
large data-sets/
persistence and querying

4. Introduction:
New Class of DBMS
4
Meta-Models
Models
Schemas
XML
Gammars
Code
Classes
Objects
ER-Schemas
Relational Data
*
-
Big Data
+
-
Graphs
ER-Schemas
Big Relational Data
?

5. Representation: Strategies
5
Object-by-object Fragments
Part-of-source Morsa, (Java) XMI, EMF-Frag
Relations CDO ?
References
Objects

6. Representation: Object-by-object vs. Fragmentation
(considering traversal, theoretical results)
6
10
0
10
1
10
2
10
3
10
4
10
5
10
6
10
0
10
1
10
2
10
3
10
4
10
5
Number of loaded objects [l]
no fragmentation [f=m]
optimal fragmentation
total fragmentation [f=1]
Executiontime[t](inms)
1e+00
1e+01
1e+02
1e+03
1e+04
1e+05
1e+06
Fragment size [f]

7. Representation: Object-by-object vs. Fragmentation
(considering traversal, theoretical results vs. implementation)
7
10
0
10
1
10
2
10
3
10
4
10
5
10
6
10
0
10
1
10
2
10
3
10
4
10
5
Number of loaded objects [l]
no fragmentation [f=m]
optimal fragmentation
total fragmentation [f=1]
Executiontime[t](inms)
1e+00
1e+01
1e+02
1e+03
1e+04
1e+05
1e+06
Fragment size [f]
10
0
10
1
10
2
10
3
10
4
10
5
10
6
10
0
10
1
10
2
10
3
10
4
10
5
Number of loaded objects [l]
Executiontime[t](inms)
1e+01
1e+02
1e+03
1e+04
1e+05
Fragment size [f]
optimal fragmentation

8. Representation: Object-by-object vs. Fragmentation
(considering traversal, implementation with actual model)
￭ Model traversal of Grabats models with four different sizes
and different characteristics
8
set0 set1 set2 set3 set4
0
1
2
3
4
5
6
7
8
XMI
CDO
Morsa
EMFFrag coarse
EMFFrag fine
notmeasured–extrapolated
notmeasured–extrapolated
Objectspersecond(10
4
)
set0 set1 set2 set3 set4
10
3
10
4
10
5
10
6
10
7
Numberoffragments
CDO/Morsa
EMFFrag coarse
EMFFrag fine

9. Representation: Object-by-object vs. Fragmentation
(considering query, implementation with actual model)
￭ Query of Grabats models with four different sizes and
different characteristics
9
set0 set1 set2 set3 set4
10
3
10
4
10
5
10
6
10
7
Numberoffragments
CDO/Morsa
EMFFrag coarse
EMFFrag fine
set0 set1 set2 set3 set4
0
50
100
150
200
250
300
350
Executiontime(ins)
XMI
CDO w/o SQL
CDO
Morsa w/o index
Morsa
EMFFrag coarse
EMFFrag fine
notmeasured–extrapolated
notmeasured–extrapolated
notmeasured–extrapolated
notmeasured–extrapolated

10. Representation: Part-of-source vs. Relations
(real implementation, artificial model)
10
10
0
10
2
10
4
10
6
10
1
10
2
10
3
10
4
number of outgoing references
executiontimeinms
10
0
10
2
10
4
10
6
10
1
10
2
10
3
10
4
number of outgoing references
executiontimeinms
Part of source implementation Relation implementation with individual access
access of one outgoing reference
traversal of all outgoing references
access of one outgoing reference
traversal of all outgoing references

11. Representation: Part-of-source vs. Relations
(real implementation, artificial model)
11
10
0
10
2
10
4
10
6
10
1
10
2
10
3
10
4
number of outgoing references
executiontimeinms
Part of source implementation
access of one outgoing reference
traversal of all outgoing references
10
0
10
2
10
4
10
6
10
1
10
2
10
3
10
4
number of outgoing references
executiontimeinms
Relation implementation with scanning
access of one outgoing reference
traversal of all outgoing references

12. 1
2
3
4
Implementation: EMF-Fragments
12
map/reduce
(hadoop)
“Share Nothing” Nodes
(cluster, adhoc-network)
DFS
(HDFS)
key-value-store
(hbase)
structured datadata-sets
applications meta-model
structured datamodel transformations

13. Implementation: Datastore mapping
13
regular containment
metamodel
0
1
part of source fragmentation
relation based fragmentation

14. Implementation: Meta-mode-based
declaration of representations
14
Project
Package
CompilationUnit
FieldMethod
Class
«fragments»
«fragments»
«fragments»
*
*
*
*
*
*
Call
«relation»

15. Implementation: Architecture
15
FragmentedModel extends Resource
ResourceSet
FObject extends EObject
FStore extends EStore
ResourceSet
Fragment extends Resource
FInternalObject
extends DynamicEObject
URIHandler
DataStore
*
*
1
*
*
1
11
1
visibleAPI
EMF-Fragments Classes
Regular EMF Classes
1
EList
EObjectEList FValueSetList
*
1
*

16. Applications: Mining and Analyzing Software
Repositories
￭ Software repositories contain more information than the current
software code:
￭ “developers who changed class/method/statement X also changed class/
method/statement Y”
￭ this information leads to knowledge about dependencies that cannot be
determined through static or even dynamic analysis
￭ this can be used to
• predict/find bugs
• understand/improve the code-base
￭ dependency information should be stored as relational data
￭ When a piece of software evolves, its metrics change. Such
dynamic metrics describe software better than static code metrics.
Could lead to a better assessment of methodologies or
understanding of software engineering in general.
16

17. Applications: Mining and Analyzing Software
Repositories
￭ JGit: Java implementation of the Git version control system
￭ MoDisco: Reverse engineering framework for eclipse java
projects based on EMF
￭ EMF-Compare: Determines matches and differences between
models
￭ EMF-Fragments: My own framework for large models
￭ over 300 Git repositories with eclipse plug-ins that
constitute the whole eclipse foundation source base as
“example” data-set
17

18. Applications: Model of a Software Repository
18
A B C
A
A B
A D
PB1.R1
B1.R2
B1.R3
B1.R4
B2.R1
B2.R2
A
A B
Repository
Revision Diff
Compilation
Unit
Model
Package Class
...
* * * *
*
1
prevnext
JGit MoDisco
modelmetamodel
usageIn
Package
Access
*
package1
«relation,
fragmentation»
«fragmentation» «relation,
fragmentation»
«relation»
«fragmentation»
* *
extends1

19. Summary
￭ Choosing the right representation makes a difference
￭ Meta-model-based declaration of representations works
(might not be good enough)
￭ There are applications that can benefit from different
representations
19
Object-by-object Fragments
Part-of-source Morsa, (Java) XMI, EMF-Frag
Relations CDO ?
References
Objects

20. Backup
20

21. Possible Approaches: Different Target Platforms
21
Schemas
XML
*
-
Big Data
-
Graphs
BASE
CAP-Theorem1
1Eric A. Brewer: Towards robust distributed systems; 19th ACM Symposium on Principles of Distributed Computing, 2000
2K. Barmpis and D.S. Kolovos. Comparative Analysis of Data Persistence Technologies for Large-Scale Models. XM 2012
ORM
XMI
XM
I+Resources
ER-Schemas
Relational Data
ACID,
structured data
ER-Schemas
Big Relational Data
BASE,
structured data
BASE,
structured data
Big
*
ORM?
2

22. Possible Approaches: Different Types of
Mapping
22
*
1Javier Espinazo-Pagán, Jesús Sánchez Cuadrado, Jesús García Molina: Morsa, A Scalable Approach for Persisting and Accessing Large Models; MoDELS 2011
per object m
apping fragm
entation
ER-Schemas
Relational Data
fast query,
slow traversal,
slow entry,
(fine transactions)
fast query,
slow traversal,
slow entry,
(fine transactions)1
Big
*
perobject
mapping
slow query,
fast traversal,
fast entry,
(coarse trans.)
Big
*ER-Schemas
Big Relational Data
/

23. Fragmentation: Types of references
￭ organizing large artifacts in different resources is already
implemented in EMF
￭ resources are loaded if necessary, objects in unloaded
resources are represented by proxy objects
￭ objects in different resources (as all related objects) are
related through references, therefore models are
fragmented along references
￭ EMF-Fragments automatically fragments large models based
on annotations in the meta-model
￭ resources are identified via URIs and can be serialized (e.g.
XMI), therefore resources can be stored in a key-value store
23

24. Fragmentation: Types of references
24
*
normal
references
*
«fragments»fragmenting
references
large value
sets *

25. Applications
￭ HWL sensor and network operation data (or experiment data in general)
￭ realtime persistence required ➜ fast data entry
￭ hierarchical structured data (different sensors and other data sources) ➜ meta-modeling
￭ queries for experiments, sensors, specific time periods ➜ only coarse simple queries
￭ traversal of larger sub-trees, mostly applications based on data aggregation
￭ actual demand for big-data depends on size of sensor network ➜ scalability
￭ CityGML models (or geo-spatial data in general)
￭ standardized as XML-schemas ➜ XML based data
￭ special proprietary indexes (e.g. spacial indexes like R-trees) and corresponding queries
￭ rather query intense applications
￭ actual demand for big-data depends on LOL of the models ➜ scalability
￭ Software Engineering
￭ Code/Model Version Control
￭ Mining Software Repositories (MSR)
￭ revisions of AST-trees and differences between AST-trees ➜ existing meta-model based frameworks (e.g. designed
for reverse engineering purposes)
￭ large number of revisions causes many large value sets
￭ queries for revisions, compilation-units ➜ rather coarse queries
￭ aggregations and statistics ➜ can be expressed in an OCL-like language
￭ immediate demand for processing in (at least smaller) clusters
￭ has to be mixed with relational data for some applications
25

26. Applications: Scientific Data
26
WSN
<xml?...>
<xml?...>
click
*
*
xml-to-model
text-to-model*

27. Applications: CityGML
￭ XML-based standard ➜ meta-models can be generated (1-
to-1 mapping)
￭ different standards define XML-schemas that extend each
other: GML⇽CityGML⇽extensions
￭ transparent use of spacial indexes
￭ map onto existing platforms (e.g. SpatialHadoop)
￭ use existing implementations and persist into the key-value
store
￭ extensions to CityGML can be facilitated to reference
CityGML-models as spatial context for sensor data
27

28. backup
28

29. Research Overview
29
W
IRELESS SENSOR NETWORKS
DATA
ANALYSISFRAMEWORK
G
EO
INFORMATION SYSTEMS
sensor data
heterogenous networks
mesh-
networks
cellular-
networks
spatial data
regular databases
spatial databases
distributed
data stores
distributed
analysis
data homo-
genisation
domain
specific
analysis
languages

30. HWL: Commodity Hardware
30

31. 31

32. ‣120+ Nodes
‣indoor and outdoor
‣dense and sparse
‣short and long links
‣stationary and mobil nodes

33. ‣120+ Nodes
‣indoor and outdoor
‣dense and sparse
‣short and long links
‣stationary and mobil nodes

35. 1
2
3
4
6
7
8
9
stein
? m
10m
5 10
Richtung Groß-Berliner
Damm
Richtung Institut
MarkusScheidgen:HWL–AHigh-PerformanceWirelessSensorResearchNetwork
35
Experiments: The Test Site
§ simplest case: two lane,
newly paved road
§ spatially equally distributed
nodes on both sides of the
rode
§ 2x5 nodes
§ homogeneous test-bed:
same nodes, equally
calibrated, same stone
ground
§ one camera to record control
data

36. 0 20 40 60 80 100 120 140 160 180 200
0
50
100
150
200
250
300
350
400
450
Single−sided Amplitude Spectrum
Frequency (Hz)
|Y(fr)|
Channel Z
Channel Y
Channel X
0 500 1000 1500 2000 2500 3000
−2
−1.5
−1
−0.5
0
0.5
1
1.5
2
Time sample (1/400 sec)
Acceleratorvalue
Time signal of all 3 channels
Channel Z
Channel Y
Channel X
MarkusScheidgen:HWL–AHigh-PerformanceWirelessSensorResearchNetwork
Experiments: Example Data
36
Amplitudes Frequencies

38. MarkusScheidgen:HWL–AHigh-PerformanceWirelessSensorResearchNetwork
Experiment: Algorithm
§ Similar to earthquake detection: comparison of
short and long moving averages (S=0.2s, L=4s)
38
sx = xth acceleration value (1)
mavg(sx, W) =
Px
i=x W si
W
(2)
ˆsx = |sx avg(sx, L)| (3)
wS
x = mavg(ˆsx, S) (4)
wL
x = mavg(ˆsx, L) (5)
w = wS
x wL
x (6)

39. Data Management
39

40. Research Overview
40
W
IRELESS SENSOR NETWORKS
DATA
ANALYSISFRAMEWORK
G
EO
INFORMATION SYSTEMS
sensor data
heterogenous networks
mesh-
networks
cellular-
networks
spatial data
regular databases
spatial databases
distributed
data stores
distributed
analysis
data homo-
genisation
domain
specific
analysis
languages

41. 41
internet
cellular
cellular
wifi
zigbee
zigbee
Technological Infrastructure

42. Logical Infrastructure
actions
visualization
sensors
information

43. 43
internet
cellular
cellular
wifi
zigbee
zigbee
information/knowledge
distributed programming
models
data bases
data representation
algorithms
processes
programming languages
CPUs
machine code radios
network protocols
hard drives
genericdomainspecific
software engineering
algorithms
processes
programming languages
information/knowledge
distributed programming
models
data bases
data representation
DSL

44. Complex Data Types
44
➡ complex data structures
➡ lots of links between data objects
➡ evolving structures
➡ requires a type safe programming
environment that proliferates re-
use

45. Large Amounts of Data
45
➡ a certain amount of data needs to be
stored per second (HWL: 120 nodes)
~140x103 data objects per second
~7MB/s serialized
➡ a certain amount of data needs to be
stored all together (24h)
~12x109 data objects
~600GB serialized
➡ Data analysis must complete in
reasonable time. For live
applications in real time.

46. From Click to ClickWatch
46
Click API software
Element
Element
Element
Compound
Handler
Handler
NetworkInterface

47. Complex Data Types: Meta-Modeling
47
This [ ] happens all the time in software modeling
state charts class diagrams MSCsOCL
context Foo
self.properties->
foreach(a|a.x != a.y)
eclipse modeling framework (EMF)
➡ Distributed storage and links between different types of data is only a simple
extension of existing technology: multi resource persistence is already implemented

48. “Share Nothing” Nodes
(cluster, adhoc-network)
DFS
(HDFS)
key-value-store1
(hbase)
Large Amounts of Data:
Problem Statement
48
1. Fay Chang, Jeffrey Dean, Sanjay Ghemawat, Wilson C. Hsieh, Deborah A. Wallach, Michael Burrows, Tushar
Chandra, Andrew Fikes, and Robert Gruber. Bigtable: A distributed storage system for structured data (awarded
best paper!). In Brian N. Bershad and Jeffrey C. Mogul, editors, OSDI, pages 205–218. USENIX Association, 2006.
2. Jeffrey Dean and Sanjay Ghemawat. Map/reduce: Simplified data processing on large clusters. In OSDI, pages 137–
150. USENIX Association, 2004.
map/reduce2
(hadoop)
hierarchical data
(XML, OGC standards)
data series
(sensor data)
signal analysis, statistics, sensor-fusion
domainspecificgeneric

49. 1
2
3
4
Large Amounts of Data: Approach
49
map/reduce
(hadoop)
“Share Nothing” Nodes
(cluster, adhoc-network)
DFS
(HDFS)
key-value-store
(hbase)
hierarchical data
(XML, OGC standards)
data series
(sensor data)
signal analysis, statistics, sensor-fusion meta-model
structured datamodel transformations