Presentation

Introduction
Related Works
Model
Experiments
Conclusion
References
C2LP: Modelling Load Propagation and

Evaluation through the Cloud Components
Rafael Mendes1 , Rafael Uriarte2 , Carlos Westphall1
1 Department
of Informatics and Statistics (INE)

Federal University of Santa Catarina (UFSC)
2 System Modelling and Analysis (SysMA)
School for Advanced Studies Lucca (IMT)
Feb 2016 / ICN2016
Rafael Mendes, Rafael Uriarte, Carlos Westphall

C2LP: Modelling Load Propagation and Evaluation through the Cloud Components
UFSC & IMT
Introduction
Related Works
Model
Experiments
Conclusion
References
Introduction
Cloud Computing Management Needs
As a Service computing offered by Cloud Computing (CC) requires

new method of management.

UFSC & IMT
Introduction
Related Works
Model
Experiments
Conclusion
References
Introduction
Challenges for CC management
The management of CC needs to deal with:

big number of variables to monitor;
several criteria to be satisfied, for multiple stakeholders and
distinct aspects as: performance, security, accounting ...;
many actions/configurations to be performed in several
abstraction levels (from services to resources, passing through
middlewares);
complex behaviour of managed elements due systems
composition and dynamism.

UFSC & IMT
Introduction
Related Works
Model
Experiments
Conclusion
References
Introduction
Where are the tools?
Decision theory [1] and Managerial Science [2] have lots of

consolidated formal methods to manage decision-making.

UFSC & IMT
Introduction
Related Works
Model
Experiments
Conclusion
References
Introduction
Where are the lack?
Based on our previous studies [3], [4], [5] and [6], we concluded
that CC lacks a managed elements model to formally represents
the load propagation through the several abstraction levels of the
cloud.

UFSC & IMT
Introduction
Related Works
Model
Experiments
Conclusion
References
Works and criteria

Ten works which present models to CC elements were analysed
according five criteria, described as follow:
Class defines if the model is for general use or for specific
services or cloud instances;
Formalism depicts if the model is based on mathematical
formalisms;
Components is related to the models capacity to express
cloud components in different abstraction levels (from services
to resources);
Comparison evaluate the capacity of the model to compare
clouds;
Simulation explain if the model has capacity to simulate
behaviours.
UFSC & IMT
Introduction
Related Works
Model
Experiments
Conclusion
References
Survey
Evaluating the related works
Table 1 : Comparison between related models. represents a feature,

a partially covered one and - when the feature is not supported.
Model
CloudSim [13]
GreenCloud [14]
iCanCloud [15]
EMUSIM [16]
MDCSim [17]
Chang[12]
P
uschel [11]
Nesmachnow [10]
Silva. [8]
Vilaplana [7]
C2LP
Class
General
General
General
General
General
Specific
Specific
Specific
Specific
Specific
General
Formalism

Components


Comparison

Simulation

UFSC & IMT
Introduction
Related Works
Model
Experiments
Conclusion
References
Model
Using multgraph for load propagation modelling
We used Direct Acyclic Multigraph (more than one edge between

two nodes), with weights in edges and nodes, to model the load
propagations through the system.

UFSC & IMT
Introduction
Related Works
Model
Experiments
Conclusion
References
Model
Graphic example
Figure 1 : Example of the propagation of loads and the evaluation processes using the
C2LP model.
UFSC & IMT
Introduction
Related Works
Model
Experiments
Conclusion
References
Elements
The mean of the weights
Edge-weights model loads. wvi represents the tuple with all

incoming edge-weight for a node v .
Node-weights model the state of node. Node-weight are tuples
composed by a priori configurations and a posteriori evaluations
(c1 , . . . , cn , gn+1 , . . . , gn+m ).

UFSC & IMT
Introduction
Related Works
Model
Experiments
Conclusion
References
Load propagation
How the loads are modelled and propagated
Edges weights represents loads and are propagated using functions

fv : Nn , Ri Ro (i.e. load balancer propagation function:
f LB
0 )).
(c1 , . . . , c5 , w1 , , w5 ) 7 (w10 , . . . , w14
edge-weight
edge-weight
edge-weight
edge-weight
(node-weight)
f(...)
edge-weight
edge-weight
edge-weight
edge-weight
Figure 2 : For load propagation have been used a breadth-first algorithm.

UFSC & IMT
Introduction
Related Works
Model
Experiments
Conclusion
References
Load consumption
Load consumption and evaluation according criteria
The consumption of loads at resource nodes (leaves) and their

evaluation. A set of g functions perform the evaluations:
gn+1 (c1 , . . . , cn , wvi ), gn+2 (c1 , . . . , cn , wvi ), . . . , gn+m (c1 , . . . , cn , wvi )
edge-weight
edge-weight
edge-weight
edge-weight
(node-weight)
(g(...),...,g(...))
(node-weight')
(node-weight')
Figure 3 : Evaluation at leaf nodes.

UFSC & IMT
Introduction
Related Works
Model
Experiments
Conclusion
References
Middleware evaluation
Evaluation in non-leaf nodes
The evaluations are back propagated using a set of g 0 functions.

0
0
Described as: gn+1
(c1 , . . . , cn , wvi , wu0 v ), gn+2
(c1 , . . . , cn , wvi , wu0 v ,
0
i
0
. . . , gn+m (c1 , . . . , cn , wv , wuv )
(node-weight') (node-weight') (node-weight') (node-weight')
(node-weight')
(g'(...),...,g'(...))
(node-weight)
(node-weight') (node-weight') (node-weight') (node-weight')
Figure 4 : For non-leaf nodes evaluations have been used a depth-first algorithm.
UFSC & IMT
Introduction
Related Works
Model
Experiments
Conclusion
References
Topology
The topology of clouds instance
Computing
Service
Load
Balancer
Cloud
Manager 1
Manager
Storage
Manager
Server 1
Platform
11
Platform
12
Platform
13
Work
Server 11
Work
Server 12
Work
Server 13
Cloud
Manager 2
Manager
Server 2
Platform
23
Platform
22
Platform
21
Work
Server 23
Work
Server 22
Work
Server 21
Work
Storage
Phisical
Network
Figure 5 : Cloud instance topology modelled for numerical experiments.

UFSC & IMT
Introduction
Related Works
Model
Experiments
Conclusion
References
Types of load
From operations in the services to loads in the elements
The service offered by the cloud supports 5 types of operations:

deploy VM,
undeploy VM,
start VM,
stop VM,
and execute tasks

UFSC & IMT
Introduction
Related Works
Model
Experiments
Conclusion
References
Functions
Fulfilling the topology with functions
4 types of propagation functions one for the service and tree

for each type of system;
6 types of leaf evaluation functions two specific performance
evaluations, one for computing resources and another for
storage and networking; plus, four common evaluation
functions (availability, accounting, security and CO2
emissions) for each type of resource;
5 types of non-leaf evaluations functions.

UFSC & IMT
Introduction
Related Works
Model
Experiments
Conclusion
References
Topology combinatorics
Variations in evaluation functions at leaf nodes
Two variations for each evaluation function (five) in leaf nodes

result in 255 cloud instance variations.
Computing
Service
Load
Balancer
Cloud
Manager 1
Platform
11
Platform
12
Platform
13
Work
Server 11
Work
Server 12
Work
Server 13
Cloud
Manager 2
Manager
Server 1
Manager
Storage
Manager
Server 2
25
25
25
Platform
23
Platform
22
Platform
21
Work
Server 23
Work
Server 22
Work
Server 21
Work
Storage
25
25
25
25
Phisical
Network
25
25
25
25
Figure 6 : Two variations for each five evaluation function in leaf nodes.
UFSC & IMT
Introduction
Related Works
Model
Experiments
Conclusion
References
Configurations
To operations to loads
The configuration determines how the 5 types of load (1 deploy

VM; 2 undeploy VM; 3 start VM; 4 stop VM; 5 and
execute tasks) will be propagated till the platforms.
This results in 52 possible different configuration of propagations.

UFSC & IMT
Introduction
Related Works
Model
Experiments
Conclusion
References
Configuration combinatorics
Propagating loads withou isomorphisms
123 4 5
Computing
Service
Load
Balancer
Cloud
Manager 1
Manager
Storage
Manager
Server 1
1235
Platform
11
Platform
12
Platform
13
Work
Server 11
Work
Server 12
Work
Server 13
Cloud
Manager 2
Manager
Server 2
Platform
23
Platform
22
Platform
21
Work
Server 23
Work
Server 22
Work
Server 21
Work
Storage
Phisical
Network
11121
Figure 7 : The load propagation using partition set notation and avoiding
isomorphisms.
UFSC & IMT
Introduction
Related Works
Model
Experiments
Conclusion
References
Evaluation
u = (av + s (p + ac + c))
(1)
where:
p average time to resolve the loads;
av average availability (in %);
ac average accounting (in currency units);
s average security (in % of risk of data exposition);
c average CO2 (in grammes).

UFSC & IMT
Introduction
Related Works
Model
Experiments
Conclusion
References
Simulation and Results
Table 2 : Summary of average evaluations for each configuration.
Criteria
Code
Time
Availability
Accounting
Security
Emissions
Utility
11221
180.59976
0.9979606
78.69924
0.9979606
82848.31
1.0526400204
Configuration
11231
11232
180.5999
180.60004
0.99795955
0.9979587
78.69926
78.699234
0.99795955
0.9979587
82848.14
82848.51
1.0526410547 1.0526477776

11212
180.59991
0.99795926
78.699265
0.99795926
82848.74
1.0526491889
UFSC & IMT
Introduction
Related Works
Model
Experiments
Conclusion
References
Conclusion
C2 LP is a flexible well-formed modelling tool to express flow of

loads through the cloud components.
This model supports the specification of elements in distinct
abstraction levels, the generation of combinatorial variations in a
use case modelling and the evaluation of the consequences of
different configuration in the load propagation.

UFSC & IMT
Finish
Thank You!
Introduction
Related Works
Model
Experiments
Conclusion
References
Itzhak Gilboa,
Theory of Decision under Uncertainty,
Cambridge University Press, 2009.
Cliff Ragsdale,
Modeling & Decision Analysis,
Thomson, 2008.
Rafael Mendes et al.,
Decision-theoretic planning for cloud computing,
In ICN 2014, The Thirteenth International Conference on
Networks, Iaria, vol. 7, no. 3 & 4, 2014, pages 191197.

UFSC & IMT
Introduction
Related Works
Model
Experiments
Conclusion
References
Alexandre A. Flores, Rafael de S. Mendes, Gabriel B. Brascher,

Carlos B. Westphall, and Maria E. Villareal,
Decision-theoretic model to support autonomic cloud
computing,
In ICN 2015, The Fourteenth International Conference on
Networks, Iaria, vol. 8, no. 1 & 2, 2015, pages 218223.
Rafael Brundo Uriarte,
Supporting Autonomic Management of Clouds:
Service-Level-Agreement, Cloud Monitoring and Similarity
Learning,
PhD thesis, IMT Lucca, 2015.

UFSC & IMT
Introduction
Related Works
Model
Experiments
Conclusion
References
Rafael Brundo Uriarte, Francesco Tiezzi, and Rocco De

Nicola,
SLAC: A formal service-level-agreement language for cloud
computing,
In Proceedings of the 2014 IEEE/ACM 7th International
Conference on Utility and Cloud Computing, IEEE Computer
Society, 2014, pages 419426.
Jordi Vilaplana, Francesc Solsona, and Ivan Teixido,
A performance model for scalable cloud computing,
In 13th Australasian Symposium on Parallel and Distributed
Computing (AusPDC 2015), ACS, vol. 163, 2015, pages
5160.
Paulo F Silva, Carlos B Westphall, and Carla M Westphall,
Model for cloud computing risk analysis,
ICN 2015, Iaria, vol. 8, no. 1 & 2, 2015, page 152.
UFSC & IMT
Introduction
Related Works
Model
Experiments
Conclusion
References
Nada Ahmed and Ajith Abraham,

Modeling cloud computing risk assessment using machine
learning,
In Afro-European Conference for Industrial Advancement,
Springer, 2015, pages 315325.
Sergio Nesmachnow, Santiago Iturriaga, and Bernabe
Dorronsoro,
Efficient heuristics for profit optimization of virtual cloud
brokers,
Computational Intelligence Magazine, IEEE, vol. 10, no. 1,
2015, pages 3343.

UFSC & IMT
Introduction
Related Works
Model
Experiments
Conclusion
References
Tim P
uschel, Guido Schryen, Diana Hristova, and Dirk
Neumann,
Revenue management for cloud computing providers:
Decision models for service admission control under
non-probabilistic uncertainty,
European Journal of Operational Research, Elsevier, vol. 244,
no. 2, 2015, pages 637647.
Chunling Cheng, Jun Li, and Ying Wang,
An energy-saving task scheduling strategy based on vacation
queuing theory in cloud computing,
Tsinghua Science and Technology, IEEE, vol. 20, no. 1, 2015,
pages 2839.

UFSC & IMT
Introduction
Related Works
Model
Experiments
Conclusion
References
Rodrigo N. Calheiros, Rajiv Ranjan, Anton Beloglazov, Csar

A. F. De Rose, and Rajkumar Buyya,
Cloudsim: a toolkit for modeling and simulation of cloud
computing environments and evaluation of resource
provisioning algorithms,
Software: Practice and Experience, Wiley Online Library, vol.
41, no. 1, 2011, pages 2350.
Dzmitry Kliazovich, Pascal Bouvry, and Samee Ullah Khan,
GreenCloud: a packet-level simulator of energy-aware cloud
computing data centers,
The Journal of Supercomputing, Springer, 2012, page
12631283.

UFSC & IMT
Introduction
Related Works
Model
Experiments
Conclusion
References
Alberto N
un
ez et al.,
iCanCloud: A flexible and scalable cloud infrastructure
simulator,
Journal of Grid Computing, Springer, vol. 10, no. 1, 2012,
pages 185209.
Rodrigo N Calheiros, Marco AS Netto, Cesar AF De Rose, and
Rajkumar Buyya,
EMUSIM: an integrated emulation and simulation
environment for modeling, evaluation, and validation of
performance of cloud computing applications,
Software: Practice and Experience, Wiley Online Library, vol.
43, no. 5, 2013, pages 595612.

UFSC & IMT
Introduction
Related Works
Model
Experiments
Conclusion
References
Seung-Hwan Lim, Bikash Sharma, Gunwoo Nam, Eun Kyoung

Kim, and Chita R Das,
MDCSim: A multi-tier data center simulation, platform,
In Cluster Computing and Workshops, 2009. CLUSTER09.
IEEE International Conference on, IEEE, 2009, pages 19.
Toufik Mansour,
Combinatorics of Set Partitions,
CRC Press, 2012.

UFSC & IMT

Presentation

Transféré par

Informations du document

Copyright

Formats disponibles

Partager ce document

Partager ou intégrer le document

Options de partage

Avez-vous trouvé ce document utile ?

Ce contenu est-il inapproprié ?

Droits d'auteur :

Formats disponibles

Presentation

Transféré par

Droits d'auteur :

Formats disponibles

Introduction

C2LP: Modelling Load Propagation and

of Informatics and Statistics (INE)

Feb 2016 / ICN2016

Rafael Mendes, Rafael Uriarte, Carlos Westphall

UFSC & IMT

As a Service computing offered by Cloud Computing (CC) requires

Rafael Mendes, Rafael Uriarte, Carlos Westphall

UFSC & IMT

The management of CC needs to deal with:

Rafael Mendes, Rafael Uriarte, Carlos Westphall

UFSC & IMT

Decision theory [1] and Managerial Science [2] have lots of

Rafael Mendes, Rafael Uriarte, Carlos Westphall

UFSC & IMT

Rafael Mendes, Rafael Uriarte, Carlos Westphall

UFSC & IMT

Works and criteria

UFSC & IMT

Table 1 : Comparison between related models. represents a feature,

Rafael Mendes, Rafael Uriarte, Carlos Westphall

We used Direct Acyclic Multigraph (more than one edge between

Rafael Mendes, Rafael Uriarte, Carlos Westphall

UFSC & IMT

UFSC & IMT

Edge-weights model loads. wvi represents the tuple with all

Rafael Mendes, Rafael Uriarte, Carlos Westphall

UFSC & IMT

Edges weights represents loads and are propagated using functions

Figure 2 : For load propagation have been used a breadth-first algorithm.

UFSC & IMT

The consumption of loads at resource nodes (leaves) and their

Figure 3 : Evaluation at leaf nodes.

Rafael Mendes, Rafael Uriarte, Carlos Westphall

UFSC & IMT

The evaluations are back propagated using a set of g 0 functions.

(node-weight') (node-weight') (node-weight') (node-weight')

UFSC & IMT

Figure 5 : Cloud instance topology modelled for numerical experiments.

Rafael Mendes, Rafael Uriarte, Carlos Westphall

UFSC & IMT

The service offered by the cloud supports 5 types of operations:

Rafael Mendes, Rafael Uriarte, Carlos Westphall

UFSC & IMT

4 types of propagation functions one for the service and tree

Rafael Mendes, Rafael Uriarte, Carlos Westphall

UFSC & IMT

Two variations for each evaluation function (five) in leaf nodes

UFSC & IMT

The configuration determines how the 5 types of load (1 deploy

Rafael Mendes, Rafael Uriarte, Carlos Westphall

UFSC & IMT

UFSC & IMT

Rafael Mendes, Rafael Uriarte, Carlos Westphall

UFSC & IMT

Simulation and Results

Table 2 : Summary of average evaluations for each configuration.

Rafael Mendes, Rafael Uriarte, Carlos Westphall

UFSC & IMT

C2 LP is a flexible well-formed modelling tool to express flow of

Rafael Mendes, Rafael Uriarte, Carlos Westphall

UFSC & IMT

Rafael Mendes, Rafael Uriarte, Carlos Westphall

UFSC & IMT