Professional Documents
Culture Documents
Textbook Hybrid Intelligent Systems Ajith Abraham Ebook All Chapter PDF
Textbook Hybrid Intelligent Systems Ajith Abraham Ebook All Chapter PDF
Abraham
Visit to download the full and correct content document:
https://textbookfull.com/product/hybrid-intelligent-systems-ajith-abraham/
More products digital (pdf, epub, mobi) instant
download maybe you interests ...
https://textbookfull.com/product/intelligent-systems-design-and-
applications-ajith-abraham/
https://textbookfull.com/product/hybrid-intelligent-systems-19th-
international-conference-on-hybrid-intelligent-systems-
his-2019-held-in-bhopal-india-december-10-12-2019-ajith-abraham/
https://textbookfull.com/product/proceedings-of-the-16th-
international-conference-on-hybrid-intelligent-systems-
his-2016-1st-edition-ajith-abraham/
Intelligent Systems Design and Applications 19th
International Conference on Intelligent Systems Design
and Applications ISDA 2019 held December 3 5 2019 Ajith
Abraham
https://textbookfull.com/product/intelligent-systems-design-and-
applications-19th-international-conference-on-intelligent-
systems-design-and-applications-isda-2019-held-
december-3-5-2019-ajith-abraham/
https://textbookfull.com/product/proceedings-of-the-third-
international-scientific-conference-intelligent-information-
technologies-for-industry-iiti18-volume-1-ajith-abraham/
https://textbookfull.com/product/proceedings-of-the-third-
international-scientific-conference-intelligent-information-
technologies-for-industry-iiti18-volume-2-ajith-abraham/
Advances in Intelligent Systems and Computing 734
Hybrid
Intelligent
Systems
17th International Conference on
Hybrid Intelligent Systems (HIS 2017)
Held in Delhi, India, December 14–16,
2017
Advances in Intelligent Systems and Computing
Volume 734
Series editor
Janusz Kacprzyk, Polish Academy of Sciences, Warsaw, Poland
e-mail: kacprzyk@ibspan.waw.pl
The series “Advances in Intelligent Systems and Computing” contains publications on theory,
applications, and design methods of Intelligent Systems and Intelligent Computing. Virtually
all disciplines such as engineering, natural sciences, computer and information science, ICT,
economics, business, e-commerce, environment, healthcare, life science are covered. The list
of topics spans all the areas of modern intelligent systems and computing.
The publications within “Advances in Intelligent Systems and Computing” are primarily
textbooks and proceedings of important conferences, symposia and congresses. They cover
significant recent developments in the field, both of a foundational and applicable character.
An important characteristic feature of the series is the short publication time and world-wide
distribution. This permits a rapid and broad dissemination of research results.
Advisory Board
Chairman
Nikhil R. Pal, Indian Statistical Institute, Kolkata, India
e-mail: nikhil@isical.ac.in
Members
Rafael Bello Perez, Universidad Central “Marta Abreu” de Las Villas, Santa Clara, Cuba
e-mail: rbellop@uclv.edu.cu
Emilio S. Corchado, University of Salamanca, Salamanca, Spain
e-mail: escorchado@usal.es
Hani Hagras, University of Essex, Colchester, UK
e-mail: hani@essex.ac.uk
László T. Kóczy, Széchenyi István University, Győr, Hungary
e-mail: koczy@sze.hu
Vladik Kreinovich, University of Texas at El Paso, El Paso, USA
e-mail: vladik@utep.edu
Chin-Teng Lin, National Chiao Tung University, Hsinchu, Taiwan
e-mail: ctlin@mail.nctu.edu.tw
Jie Lu, University of Technology, Sydney, Australia
e-mail: Jie.Lu@uts.edu.au
Patricia Melin, Tijuana Institute of Technology, Tijuana, Mexico
e-mail: epmelin@hafsamx.org
Nadia Nedjah, State University of Rio de Janeiro, Rio de Janeiro, Brazil
e-mail: nadia@eng.uerj.br
Ngoc Thanh Nguyen, Wroclaw University of Technology, Wroclaw, Poland
e-mail: Ngoc-Thanh.Nguyen@pwr.edu.pl
Jun Wang, The Chinese University of Hong Kong, Shatin, Hong Kong
e-mail: jwang@mae.cuhk.edu.hk
Editors
123
Editors
Ajith Abraham Azah Kamilah Muda
Machine Intelligence Research Labs Faculty of Information and Communication
Auburn, WA Technology
USA Universiti Teknikal Malaysia Melaka
Durian Tunggal, Melaka
Pranab Kr. Muhuri Malaysia
Department of Computer Science
South Asian University Niketa Gandhi
Chanakyapuri, Delhi Machine Intelligence Research Labs
India Auburn, WA
USA
This Springer imprint is published by the registered company Springer International Publishing AG
part of Springer Nature
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
Preface
v
vi Preface
intelligence areas. HIS 2017 received submissions from over 24 countries, and each
paper was reviewed by at least five reviewers in a standard peer review process.
Based on the recommendation by five independent referees, finally 35 papers were
accepted for publication in the proceedings published by Springer, Verlag.
We would like to thank the Springer publication team for the wonderful support
for the publication of this volume.
Ajith Abraham
Pranab Kr. Muhuri
HIS 2017 General Chairs
HIS 2017 Organization
General Chairs
Advisory Board
vii
viii HIS 2017 Organization
Publication Chairs
Web Service
Additional Reviewers
xi
xii Contents
1 Introduction
Uniaxial compressive strength (UCS) is the most familiar parameter used in rock
engineering projects. In rock mass rating, proposed by Bieniawski [1], UCS is the
only parameter to assess the strength of rock material. However, high quality
machined specimens are required for determining UCS in the laboratory [2].
This makes the test expensive and time-consuming. Therefore, different empirical
predictive models are being used to estimate the UCS of rock materials indirectly
from the index tests, which require little or no specimen preparation and are
less expensive than the uniaxial compression test. Amongst different predictive
models, regression analysis is the most commonly used. Lists of such relationships
between the index tests and UCS obtained by different researchers have been
given in [3].
c Springer International Publishing AG, part of Springer Nature 2018
A. Abraham et al. (Eds.): HIS 2017, AISC 734, pp. 1–10, 2018.
https://doi.org/10.1007/978-3-319-76351-4_1
2 V. K. Ojha and D. A. Mishra
In the last decade, however, the use of computational intelligence (CI) meth-
ods to establish predictive models has gained much interest in the areas of rock
mechanics. A comprehensive list of such works is presented in Table 1. These
research works indicate the requirement of improvement in the CI methods to
estimate UCS from the index test results and to assess the effectiveness of indi-
vidual index tests for predicting the UCS of rock materials.
Although several predictive models have been proposed to estimate the UCS
from index tests, the influence of various index tests in predicting UCS is not
entirely understood. More specifically, so far, to the best of our literature sur-
vey, no study has been done on assessing the effectiveness of different index
tests. Moreover, it is always advantageous if development time and effort of the
predictive modeling can be reduced. To achieve these goals, the use of the neu-
ral network (NN) or similar CI methods on a limited set of experimental data
for predictive modeling can be a great benefit. Hence, the aim of this work is
to develop a predictive model that can estimate the UCS with a high accuracy
and assess the effectiveness of different index tests in predicting UCS of rock
materials.
The multiobjective heterogeneous flexible neural tree (HFNTM ) produces a
tree-like model [17], in which the nodes of the tree are similar to the NN nodes
was used in this study. Interestingly, HFNTM differs from NN [18] in its struc-
tural configuration, and it differs from the commonly used regression tree [19]
in its node type. Moreover, the tree-like structure in the HFNTM is created by
using multiobjective genetic programming (MOGP) [17,20]. Therefore, the pri-
mary advantages of using HFNTM over other CI techniques lie in its ability of
the automatic adaptation into the structure and the input feature selection.
Such an ability like HFNTM is necessary and has been successfully used in
several real-life applications [21]. However, this method has not been exploited
in estimating the UCS from index tests. In this work, index tests namely
block punch index (BPI), point load strength (Is-50), Schmidt rebound hard-
ness (SRH), ultrasonic P-wave velocity (Vp ), and physical properties namely
effective porosity (ηe ) and density (ρ) are determined and used for estimating
the UCS of granite, schist, and sandstone. An improved CI technique, HFNTM ,
was employed for this purpose and it was used to determine the effectiveness
of index tests BPI, Is-50, SRH, Vp , ηe , and ρ in the prediction of UCS of three
rocks. Such kind of individual test effectiveness assessment using HFNTM has
not been studied in the past, and it is the noble contribution of this research
along with the predictive modeling. The evaluation of individual tests provides
a detailed insight of UCS prediction of rock materials.
same locality as this study) presented in [3]. A total 60 samples (20 from each
rock types) were used for the predictive modeling.
3 Predictive Models
3.1 Computational Intelligence Techniques
Discovering knowledge contained in data and developing predictive models are
vital tasks performed by the CI methods. Moreover, predictive modeling identi-
fies the underlying relationship between an input variable X = {x1 , x2 , . . . , xN }
and an output variable d = {d1 , d2 , . . . , dN } through the learning parameter w,
which defines the said relationship. A CI method finds the learning parameter w
by usually reducing the root mean square error (RMSE) e as per Eq. (1) between
the predicted output y = {y1 , y2 , . . . , yN } and the desired output d. Hence, the
learning of a CI method indicate the search for a proper learning parameter w.
In this study, the RMSE e was evaluated as:
1 N
e= (yi − di )2 , (1)
N i=1
where ȳ and d̄ are the average of predicted and desired outputs. The correlation
coefficient ranges from −1 to 1, where a value 1 indicates the best performance.
In this study, CI methods such as fuzzy inference system (FIS), multilayer
perceptron (MLP), and adaptive neuro-fuzzy inference system (ANFIS) were
considered apart from the proposed HFNTM . The FIS, MLP, and ANFIS are
most widely used CI methods for modeling industrial and engineering prob-
lems [22]. The brief introduction to the mentioned CI methods are as follow:
• An FIS is a rule-based system, where a set of IF-THEN rules are designed
from the given input variables. Subsequently, the rules are inferred (conclusion
drawn from the rules set) to predict the output of the system for any given
input [23].
• An MLP is an NN model, which is an imitation of human-like learning. An
MLP is a layered network of neural nodes (computational node) arranged in
a layered structure. The nodes in an MLP are connected with synaptic links.
The primary form of MLP training for a given dataset is the discovery of
appropriate values for the synaptic connections [18].
• An ANFIS is a combination both FIS and NN-based systems, where an FIS
is designed as an NN-like model. ANFIS is typically a six-layered system in
which the layers indicate the IF and THEN parts of an FIS [24].
Neural Tree for Estimating the Uniaxial Compressive Strength 5
An HFNTM similar to any other CI methods often tries to minimize the error
e given in Eq. 1 for the given dataset by optimizing its parameter using some
learning algorithms (HFNTM employ multiobjective genetic programming and
differential evolution for this purpose). The structure of HFNTM is a tree-like
that has internal nodes as the computational nodes (analogous to MLP neurons),
branches (similar to MLP synaptic links), and leaf nodes (to represent input
variables).
Mathematically, an HFNTM , denoted as G, is a union of internal node V
and the leaf node T . The internal node V is a set of internal (computational)
nodes and, the leaf node T is a set of inputs [17]. Hence, an HFNTM G can be
expressed as:
U (k) U (k) U (k)
G = V ∪ T = {v2 , v3 , . . . , vtn } ∪ {x1 , x2 , . . . , xd } (3)
U (k)
where vi (i ∈ {2, 3, . . . , tn}) indicates an internal (computational) node
that takes two or more arguments; whereas, the leaf node takes no argument.
The function U (k) randomly invoke an activation function at a computational
node from a set of activation function: {Gaussian, tangent hyperbolic, bipolar
sigmoidal, unipolar sigmoidal, and Fermi}. HFNTM training has two aspects.
Firstly, discovering appropriate tree structure and secondly, the optimization of
tree parameters. These two HFNTM training parts are performed in two phases:
where Rj is the selection rate of j-th input feature set Aj ∈ P({BPI, Is-50,
SRH, Vp , ηe , ρ }), and function I(mi = Aj ) is a function that returns “1” if
j-th input-feature set Aj is selected by the i-th model mi , otherwise, it returns
“0.” Feature selection rate Rj = 1 is the highest (i.e., all the models selected
the input-feature set (Aj ) and Rj = 0 is the lowest (i.e., no model selected the
input-feature set Aj ).
The predictability score Pj of an input feature set Aj is on the other hand is
necessary to determine along with the selection rate Rj because the models in
the list may not be equal in their performances. To determine the predictability
score Pj of j-th input-feature set Aj , the performance Fj (typically the RMSE)
of the corresponding input-feature set Aj was at first computed as per Eq. (5):
|M |
ei × I(mi = Aj ) if|Aj | = 1
Fj = i=1 |M | |M | (5)
i=1 ei × I(mi = Aj )/ i=1 I(mi = Aj ) if|Aj | > 1
where ei indicates the RMSE of i-th model. The performance Fj for |Aj | = 1 is
the sum of RMSEs and Fj for |Aj | > 1 is the average RMSEs of all models that
selected a subset Aj . Accordingly, the predictability score Pj corresponding to
an input-feature set Aj was computed by normalizing the performance as [21]:
Fj
Pj = (6)
maxj=1 to z (Fj )
where function max(.) evaluate the maximum performance value among all Fj .
Analogous to the selection rate Rj , the predictability score Pj = 1 for an input-
feature set Aj describes the heights impact on the model’s predictability and
the score Pj = 0 describes the least impact on the model’s predictability.
UCS
v32
IS-50 IS-50
v32
BPI
v21 v21
HFNTM model is better than other models. Besides the statistical parameters,
computational time of the HFNTM model was 9 s, and the complexity was 11
parameters for such a wide range of data obtained for three completely different
rock types. Considering all these factors, RMSE, r, r2 , and computation time,
it can be said the HFNTM is a very effective tool for determining the UCS
from index tests. Additionally, the HFNTM is a less complex model than the
other CI model. As shown in Table 2, the total parameters in the HFNTM were
smaller than the other CI models. Hence, HFNTM is a better model in both
implementation and computational points of view.
Table 2. Performance of the models for the selected training and test datasets.
Fig. 2. Estimated UCS plotted against measured UCS: training (left) and test (right).
of selected inputs by each model was prepared with the account of their RMSEs.
Afterward, a comprehensive feature analysis of all 6 input features was per-
formed. For the input feature analysis, two performance indicators, feature selec-
tion rate R as defined in Eq. (4) and feature predictability score P as defined in
Eq. (5), were adopted.
The effectiveness of individual index tests BPI, Is-50, SRH, Vp , ηe and ρ
in estimating the UCS were examined through R and P and setting |Aj | = 1.
Table 3 presents the results of feature analysis carried out for the all 6 individual
inputs. The input features BPI and Is-50 represent the destructive mechanical
indices; whereas, SRH and Vp represent the non-destructive rock indices. Effec-
tive porosity (ηe ) and density (ρ) are the determined physical properties of the
concerned rocks. It can be observed that R and P of destructive mechanical
indices BPI and Is-50 were much higher than that of R and P values of non-
destructive and physical rock indices.
Among the individual input features, R and P of Is-50 are 0.959 and 1 respec-
tively. Next in the list is BPI with R and P of 0.862 and 0.899 respectively. From
the performed feature analysis, it can be said that Is-50 is the best index to pre-
dict the UCS of rock materials. Since BPI also has significantly high R and P ,
we can say that destructive mechanical rock indices are the best proxy for esti-
mating the UCS of the rock material. This finding is also in accordance with the
experimental findings of [26].
Neural Tree for Estimating the Uniaxial Compressive Strength 9
5 Conclusions
The experimental program included the determination UCS, BPI, Is-50, SRH,
Vp , porosity and density of three different rock types (granite, schist, and sand-
stone). To estimate the UCS from the index tests, a multiobjective heterogeneous
flexible neural tree (HFNTM ) model was proposed. The inputs and outputs for
the developed HFNTM model were used from the rock materials experimental
results. The constructed model efficiently estimated the UCS based on the infor-
mation gathered from the experimental data in a very quick time (9 seconds).
Additionally, the HFNTM was a less complex model than the other CI model.
Therefore, this model can be effectively used in estimating the UCS. Developed
HFNTM also assess the effectiveness of different index tests in predicting UCS of
rock materials with the help of feature selection rate R and predictability score
P . From R and P values of the individual input features (index test results), it is
found that Is-50 is the best proxy for the UCS. Among the different types of index
tests—destructive indices, non-destructive indices, and physical properties—the
destructive mechanical rock indices BPI and Is-50 are found to be the best index
tests to estimate the UCS.
References
1. Bieniawski, Z.T.: Engineering Rock Mass Classifications, p. 251. Wiley, New York
(1989)
2. ISRM: The complete ISRM suggested methods for rock characterization, testing
and monitoring. In: Ulusay, R., Hudson, J.A. (eds.) Suggested Methods Prepared
by the Commission of Testing Methods, Kozan Ofset, Ankara, ISRM, 19742006.
Compilation Arranged by the ISRM Turkish National Group (2007)
3. Mishra, D.A., Basu, A.: Estimation of uniaxial compressive strength of rock mate-
rials by index tests using regression analysis and fuzzy inference system. Eng. Geol.
60, 54–68 (2013)
4. Meulenkamp, F., Grima, M.A.: Application of neural networks for the prediction
of the unconfined compressive strength (UCS) from Equotip hardness. Int. J. Rock
Mech. Min. Sci. 36, 29–39 (1999)
5. Gokceoglu, C., Zorlu, K.: A fuzzy model to predict the uniaxial compressive
strength and the modulus of elasticity of a problematic rock. Eng. Appl. Artif.
Intel. 17(1), 61–72 (2004)
6. Sonmez, H., Gokceoglu, C., Nefeslioglu, H.A., Kayabasi, A.: Estimation of rock
modulus: for intact rocks with an artificial neural network and for rock masses
with a new empirical equation. Int. J. Rock Mech. Min. Sci. 43, 224–235 (2006)
7. Zorlu, K., Gokceoglu, C., Ocakoglu, F., Nefeslioglu, H.A., Acikalin, S.: Prediction
of uniaxial compressive strength of sandstone using petrography-based models.
Eng. Geol. 96, 141–158 (2008)
8. Gokceoglu, C., Zorlu, K., Ceryanc, S., Nefeslioglu, H.A.: A comparative study on
indirect determination of degree of weathering of granites from some physical and
strength parameters by two soft computing techniques. Mater. Charact. 60, 1317–
1327 (2009)
10 V. K. Ojha and D. A. Mishra
9. Yilmaz, I., Yuksek, G.: Prediction of the strength and elasticity modulus of gypsum
using multiple regression, ANN and ANFIS models. Int. J. Rock Mech. Min. Sci.
46, 803–810 (2009)
10. Dehghan, S., Sattari, G.H., Chehreh, C.S., Aliabadi, M.A.: Prediction of uniax-
ial compressive strength and modulus of elasticity for Travertine samples using
regression and artificial neural networks. Min. Sci. Tech. 20, 41–46 (2010)
11. Rabbani, E., Sharif, F., Kooliv, M., Salooki, M.A.: Application of neural network
technique for prediction of uniaxial compressive strength using reservoir formation
properties. Int. J. Rock Mech. Min. Sci. 56, 100–111 (2012)
12. Yesiloglu-Gultekin, N., Gokceoglu, C., Sezer, E.A.: Prediction of uniaxial compres-
sive strength of granitic rocks by various nonlinear tools and comparison of their
performances. Int. J. Rock Mech. Min. Sci. 62, 113–122 (2013a)
13. Yesiloglu-Gultekin, N., Sezer, E.A., Gokceoglu, C., Bayhan, H.: An application
of adaptive neuro fuzzy inference system for estimating the uniaxial compressive
strength of certain granitic rocks from their mineral contents. Expert Syst. Appl.
40, 921–928 (2013b)
14. Armaghani, D.J., Hajihassani, M., Bejarbaneh, B.Y., Marto, A., Mohamad, E.T.:
Indirect measure of shale shear strength parameters by means of rock index tests
through an optimized artificial neural network. Measurement 55, 487–498 (2014)
15. Armaghani, D.J., Mohamad, E.T., Momeni, E., Narayanasamy, M.S., Amin,
M.F.M.: An adaptive neuro-fuzzy inference system for predicting unconfined com-
pressive strength and Youngs modulus: a study on Main Range granite. Bull. Eng.
Geol. Environ. 74, 1301–1319 (2015)
16. Mishra, D.A., Srigyan, M., Basu, A., Rokade, P.J.: Soft computing methods for
estimating the uniaxial compressive strength of intact rock from index tests. Int.
J. Rock Mech. Min. Sci. 80, 418–424 (2015)
17. Ojha, V.K., Abraham, A., Snášel, V.: Ensemble of heterogeneous flexible neural
trees using multiobjective genetic programming. Appl. Soft. Comput. 52, 909–924
(2017)
18. Haykin, S.: Neural Networks and Learning Machines, 3rd edn. Prentice Hall, Upper
Saddle River (2009)
19. Kohavi, R., Quinlan, J.R.: Data mining tasks and methods: classification: decision-
tree discovery. In: Handbook of Data Mining and Knowledge Discovery, pp. 267–
276. Oxford University Press (2002)
20. Chen, Y., Yang, B., Dong, J., Abraham, A.: Time-series forecasting using flexible
neural tree model. Inf. Sci. 174, 219–235 (2005)
21. Ojha, V.K., Schiano, S., Wu, C.Y., Snášel, V., Abraham, A.: Predictive modeling
of die filling of the pharmaceutical granules using the flexible neural tree. Neural
Comput. Appl., 1–15 (2016)
22. Lam, H.K., Nguyen, H.T.: Computational Intelligence and Its Applications: Evolu-
tionary Computation, Fuzzy Logic, Neural Network and Support Vector Machine
Techniques. World Scientific, Singapore (2012)
23. Rezaee, B., Zarandi, M.F.: Data-driven fuzzy modeling for Takagi-Sugeno-Kang
fuzzy system. Inf. Sci. 180(2), 241–255 (2010)
24. Jang, J.S.: ANFIS: adaptive-network-based fuzzy inference system. IEEE Trans.
Syst. Man Cybern. 23(3), 665–685 (1993)
25. Storn, R., Price, K.: Differential evolutiona simple and efficient heuristic for global
optimization over continuous spaces. J. Global Optim. 11, 341–359 (1997)
26. Mishra, D.A., Basu, A.: Use of the block punch test to predict the compressive
and tensile strengths of rocks. Int. J. Rock Mech. Min. Sci. 51, 119–127 (2012)
Fault Tolerant Multiple Synchronized Parallel Load
Balancing in Cloud
1 Introduction
With the emergence of cloud computing more and more business organizations are
moving towards this platform due to its low cost, simple configuration, and virtually
unlimited resource pool with on-demand provisioning. The performance of the cloud
eco system can be enhanced if the scheduling of resources is properly done. Resource
scheduling with load balancing is one of the best methods for improving the cloud
performance. The researchers have proposed several methods for optimal scheduling of
resources in the cloud.
Even though the features mentioned above are attractive, a fault tolerant mechanism
for undisrupted performance of cloud services is the most essential one. Load balancing
mechanism will improve the performance by efficient distribution of workloads across
multiple computing resources. The load balancer distribute the load evenly along cluster
members. Load balanced cluster is an abstraction for a set of identical processors, that
host same set of services. A simple cluster with a load balancer and respective cluster
members is shown in Fig. 1. Here n servers are managed by a single load balancer. When
this single load balancer fails, functioning of the entire system collapses due to the non-
availability of the load balancer cum dispatcher. This causes financial as well as credi‐
bility loss for the cloud provider.
2 Related Works
Load balancing is a cumbersome task in cloud resource allocation. The load is unpre‐
dictable in cloud computing and it may vary depending on the demand for a particular
service. Intelligent load balancing mechanisms are needed for ensuring better perform‐
ance and QoS. Load balancing algorithms are classified into static and dynamic, bio-
inspired, and game theory based. The common static algorithms are random, round
robin, min-min, min-max and weighted round robin algorithms. The paper [1] presents
a generic load balancing scheme, which separates the allocating and migrating processes
while preserving a guaranteed level of service.
An idea to make load balancing more dynamic for better QoS management of multi-
instance applications in the cloud is presented in [2]. This approach uses a load balancer
equipped with a queue for incoming user requests to limit its number, that can be effec‐
tively sent and stored in queues. Content-aware load balancing [3] is another approach.
This load balancer manages a farm of web servers and differentiates the service required
by workload and controls admissions under extreme overload. Multi-agent based distrib‐
uted load balancing framework for cloud platform [4] uses workload prediction tech‐
nology and threshold buffering strategies to reduce the migration of VMs. The paper [5]
gives an idea for accessing resources through the internet and makes use of Max-Min
and Min-Min algorithms. These algorithms are implemented based on the study of
RASA algorithm. RASA method provides better resource utilization and response time.
A fully distributed load balancing algorithm [7] is presented to cope with the load
imbalance and single point failure.
A comprehensive discussion and comparison of latest load balancing algorithms are
done in [8, 20] and these studies provides an overview of the latest approaches in this
field. Nature inspired algorithms are good candidates for solving dynamic load balancing
in cloud environment [9, 15]. Other major approaches are based on QoS [10, 11, 18,
19], Ant colony [12, 17], weighted load balancing [13], Genetic Algorithm (GA) [14]
and PSO [16] enabled load balancing methods.
3 Problem Identification
One of the main features of cloud is on-demand scalable computing at any time. In cloud,
there is no explicit knowledge for the customer about where their task is being executed
and in which server. Cloud providers are trying to offer fault tolerant services to their
customers. But the single point failures are one of the barriers for attaining fault tolerant
continuous service. Since the load balancer is responsible for distributing the tasks
received from the end users to any one of the processors, the processing of tasks will be
halted when the load balancer is down, even when the processor is ready to execute it.
This may be due to hardware failures like, server crashes, network problems, power
failures or disk crashes. Software failures like directory/proxy server crash and database
corruption also result in single point failure. So to address these failures, a suitable
cooperative mechanism is needed for fault tolerant cloud service.
14 S. Sreelekshmi and K. R. Remesh Babu
4 System Design
The proposed system contains a number of schedulers (load balancers), each capable of
balancing the task across multiple processors. These schedulers interacts each other for
gathering information about the tasks status and tasks in the input queue. They are also
able to distribute the tasks to other processors in data center based on known capability
of each processor. After the execution of each process, it can generate a feedback based
on the current capacity of each server. The capacity calculation is done in fixed time
interval based on the number of tasks executed by the processor and tasks pending in
the queue.
This frequent monitoring and cooperative load balancing ensure desired QoS to the
end users. In the proposed cooperative load balancing, none of the load balancers are
overloaded due to sharing of information about tasks already completed, being executed
or waiting in the queues. The architecture of the proposed method is shown in Fig. 2.
The detailed explanation is given in the next sub sections.
A set of tasks is given to each load balancer, which stores them in an output queue. There
exists a dispatcher for distributing the requests to different processors based on
Another random document with
no related content on Scribd:
XV
AS IT WAS IN MY DAY
I am like the deaf old lady who, when asked why she took a box at
the opera when she could not hear, replied, “I can see.” So it is on
piazzas at summer hotels, I do not overhear remarks, so perforce
the pleasure of gossip is denied me, but “I can see,” and no doubt do
observe more than those who have the other faculty to play upon;
also I see and moralize. Last summer in the mountains didn’t I see
young girls, young society girls, educated girls who ought to have
known better, with bare heads and bare arms playing tennis in the
hot sun; and, worse still, racing over the golf links? I could see them
from my window, equally exposed, chasing balls and flourishing
clubs. The sun in August is pitiless even on those breezy mountains,
so I was scarcely surprised when one young girl was overcome by
heat and exposure, and was brought to her mother at the hotel in a
passing grocer’s cart or lumber wagon. I tell my grandchildren who
want to “do like other girls” that is not the way “other girls” did in my
day. Grandma may be so old that she forgets, but she moralizes all
the same. These athletic girls come back to city homes so sunburnt
and with such coarse skin they have to repair to a skin specialist,
and have the rough cuticle burnt off with horrid acids, and be
polished up before the society season opens.
There are, of course, extremes, but years ago young ladies took
more care of their complexions and of their hair, too. Years back of
years, I don’t know how they did. In my day we girls loved to visit the
granddaughter of a voluble dame and listen to the old lady’s talk, just
like I am talking now. She thought we were criminally careless with
our “skins,” as she called it. Why, when she was young, her skin was
so thin and clear that “one saw little blue veins meandering her
neck.” We always heard something as reminiscent in that house to
laugh over till we saw the old lady again, and heard something
equally remarkable of her youth. She was living in the past, as I am
now, as I return to my experiences. One young girl visited me, ever
so many years ago, who wore one of those awful, long, scoop
sunbonnets all the time she was not at table or in bed. She looked
like the proverbial lily. I used to wish she would take off that
sunbonnet and say something, for she was dumb as a lily. I have
entirely forgotten her name, though she was my guest for a whole
stupid week; but I recall she was a relative or friend of the Morses. I
don’t know Mr. Morse’s name; he was called Guncotton Morse, for
he invented an explosive of that name, which the United States
Government appropriated during the war.
Years after this young girl’s visit to me I called on the charming
Morse family in Washington. He was then urging his “claim.” Every
Southerner in Washington was after a “claim” at that time. I nearly
broke my neck falling over a green china dog or a blue china cat in
their dark parlor. Enterprising Morse barricaded himself behind his
explosive, but I think he failed in his fight. I find I have wandered
from the girls having their skins burned off to the Morses and their
blue china cat!... In my days there were no specialists except cancer
doctors. I think they always flourished—there were no skin
specialists. A doctor was a doctor, nothing more nor less, and he
was supposed to know all that was necessary of the “human form
divine.” He did, too, for people did not have the new-fangled
diseases of to-day. A woman’s hospital! Oh, heavens! Only last week
I saw a friend, old enough to know better, but we never are so old we
don’t want to rid our faces of pimples and warts and wrinkles. This
friend was a sight. I was really alarmed for her. She had been to a
specialist. Her face was fiery red, all the skin removed by acid.
Yesterday I saw her again, cured of sunburn and all the ills skin is
heir to. Her complexion was that of the lily girl who wore the scoop
sunbonnet. I do not advise you to try the experiment. It is shockingly
painful, and does not always prove a success.
When I was a little girl, more than seventy years ago, mother
made me, for summer romps in the country, gloves of nankeen, that
well covered the wrist, had a hole for the thumb and a deep flap to
fall over the hand. It was lucky they were easily made, and nankeen
was not expensive, for I hated them and had a way of losing them in
the currant bushes. Maybe you never saw nankeen? Gentlemen’s
waistcoats were often made of it, and little boys’ trousers. If I lost my
scoop sunbonnet one day—and it was surprising how easily I lost it!
—it was sewed on the next. There were no such things as hatpins—
and we had pigtails anyway, so they would have been of no use.
Such tortures were inflicted when we were running wild over the blue
grass farm, but no doubt the little Creole girls on the lakeshore were
similarly protected. The hair specialist was not in evidence either.
Ladies had their hair done up with bandoline and pomatums made
of beef’s marrow and castor oil and scented with patchouli; hair was
done into marvelous plaits and puffs. A very much admired style
which Henriette Blondeau, the fashionable hair dresser, achieved,
was a wide plait surrounding a nest of stiff puffs. It was called the
“basket of fruit.” The front locks were tiny, fluffy curls each side the
face and long ringlets to float over the shoulders. We all remember
Henriette Blondeau. She dressed my sister’s hair in the early forties,
and she dressed mine ten years later, and I met her in the hall of the
St. Charles Hotel, plying her trade, twenty years later still, the same
Henriette, with the same ample apron, the tools of her trade sticking
out from her pockets. Now, almost forty years later still, she walks
the streets of New Orleans no more. I hope she rests somewhere in
the old French cemetery, for she knew and gossiped with so many
who are taking their long sleep in that peaceful spot.
Mother made—no doubt your grandmother did, too—the pomade
that was used on our hair. It was used, too, very freely; our locks
plastered down good and smooth and flat. You may wonder how
long hair so treated could last; just as long as hair ruffled the wrong
way and marcelled with warm irons lasts our girls to-day. Mother’s
pomade was made of beef’s marrow and castor oil. After the marrow
was rendered to a fluid state, oil was added, then perfume, the whole
beaten in a deep bowl until perfectly cold and white. Mother would
beat and beat, add a few drops more of essence of bergamot, smell
and smell and beat and smell, until she had to call a fresh nose to
see if it was all right. I remember being told to try my olfactories on
the soft, creamy stuff. A naughty brother gave my head a blow that
sent my little pug-nose to the bottom of the bowl! My face was
covered to the ears, and while mother scraped it with a spoon and
scolded Henry, she was entreating me not to cry and have tears
spoil her pomade. Maybe I might have forgotten how the stuff was
made and how it looked, but for that ridiculous prank of the dearest
brother ever was.
I have a sweet little miniature of that brother Henry, namesake of
my father’s dear friend, Henry Clay, with the queer collared coat and
flourishing necktie of the day, and his long, straight hair well
plastered with mother’s good pomade. The dear man went to Central
America, on a pleasure tour to the ruins of Uxmal in 1844. The
vessel on which he sailed for home from Campeache, in September
of that year, disappeared in the Gulf. We never had any tidings of
how, or when, or where. I remember the firm of J. W. Zacharie was
consignee of that ill-fated Doric, and how tenderly Mr. Zacharie came
to my stricken mother, and how much he did to obtain information,
and how for weeks after all hopes were abandoned my mother’s
heart refused to believe her boy was indeed lost. Every night for
months she placed with her own trembling hands a lamp in the
window of Henry’s room, to light him when he came. She never gave
up some remnant of hope. So far as I know, only one friend of that
dear brother, one contemporary, is living now, in New Orleans. She
is the last of her generation; I am the last of mine.
In those days there were few patent medicines, washes and
lotions. There was a Jayne’s hair tonic, and somebody’s chologogue,
that was a fever cure much in evidence on plantations, for quinine
and blue mass pills—others, too—were made by hand. I have made
many a pill. We had an old negro woman who was daft on the
subject of medicine. There was not an earthly thing the matter with
Hannah—she was just a chronic grumbler, begging for “any kind of
pill.” I doctored her successfully, making for her bread pills, rolling
them in a little rhubarb dust to give them a nasty taste. They did her
a world of good. Mother made our lip salve (didn’t your
grandmother?) of white wax and sweet oil. We did not have cold
cream in those days.
When by accident, or some other way, our faces tanned, a wash
overnight of sour buttermilk was all that was required. It was not very
pleasant, and nobody wanted to occupy the room with you on sour
buttermilk night. Reason obvious. Kentucky belles, who were noted
for their rosy cheeks, often increased the bloom by a brisk rubbing of
the leaves of the wild mullein. Except rice powder (and that is not a
cosmetic) no cosmetics were in use.
We can recall at a later date than my girlhood a lady from
somewhere up the coast married a finicky cotton broker in New
Orleans. They made a wedding trip to Paris, and she returned with
her face enameled. I don’t think it could have been very skillfully
done, for she had to be so careful about using the muscles of the
face that she was absolutely devoid of expression. Once, in a
moment of forgetfulness or carelessness, she “cracked a smile,”
which cracked the enamel. She returned to Paris for repairs. I saw
her on the eve of sailing, and do not know if she ever returned.
XVI
FANCY DRESS BALL AT THE MINT IN 1850
The belles retired to their boudoirs for a season, but the beaux had
to go to business, and what a sight some of them were for a whole
week after the fancy dress ball! They had hired costumes from
members of the French opera troupe, and their faces were “made
up” with rouge that could not be washed off; had to wear off in a
purplish stain. My brother represented Louis XIV on that occasion,
and I remember he scrubbed his cheeks until he made them almost
raw. Of no avail. In time the pinkish, purplish tint gradually
disappeared.
Shortly after that grandest and most unique entertainment Mr. Joe
Kennedy’s term expired and he retired into private life. Beautiful
Rose fell into a decline and died early. What fortunes befell that
family I know not. They seem to have faded away. The Kennedys
were a large family in those days, closely allied to the Pierce and
Cenas families, all of which were socially prominent. And now their
names are “writ in water.” I should like to know how many of this old
Creole society are living to-day! I was eighteen, one of the youngest
of the group, in the fifties.
XVII
DR. CLAPP’S CHURCH
(From “Forty Etchings, from Sketches Made with the Camera Lucida in North
America in 1827 and 1828,” by Captain Basil Hall, R. N.)
Grey Eagle was the finest and best, and therefore most popular
boat. I recall with amusement an eight or ten days’ trip on that
palace. The cabins were divided by curtains, drawn at night for
privacy. The ladies’ cabin, at the stern, was equipped with ten or
twelve small staterooms. The gentlemen’s cabin stretched on down
to the officers’ quarters, bar, barber shop, pantries, etc., ending in
what was called Social Hall, where the men sat about, smoking and
chewing (the latter as common a habit as cigarette smoking is now)
and talking—in other words, making themselves sociable.
On that same Grey Eagle I was for the first time promoted to the
upper berth, in a stateroom shared by an older sister. The berth was
so narrow that in attempting to turn over I fell out and landed in the
wash basin, on the opposite side of the room! My sister had to sit on
the lower berth to braid my pigtails, then sent me forth so she could
have room to braid her own. Trunks and other baggage more
unwieldy than carpetbags were piled up in the vicinity of Social Hall.
A carpetbag, small enough to be easily handled, was all there was
room for in the stateroom. There were no valises, suitcases or
steamer trunks in those days of little travel, and unless you are
three-quarters of a century old you can’t imagine a more unwieldy
article than a carpetbag of seventy years ago. Only toilet articles and
things that could not muss and tumble could be safely stored in one.
In the stateroom, where we had to sleep and dress, and, if we
could snatch a chance, take an afternoon nap, there was a corner
shelf for a basin and pitcher and one chair; two doors, one leading
out and the other leading in, transoms over each for light and
ventilation—and there you are for over a week. The cabin was
lighted with swinging whale-oil lamps, and one could light his
stateroom if one had thought to provide a candle.
Every family traveled with a man servant, whose business it was
to be constantly at beck and call. Of course, there was always a
colored chambermaid, and, equally of course, she frisked around
and seemed to have very little responsibility—no bells, no means of
summoning her from her little nodding naps if she happened to be
beyond the sound of one’s voice. The man servant’s duties,
therefore, were almost incessant. If an article was needed from the
trunks he was sent to the baggage pile for it, and often he brought
trunk trays to the staterooms. When the boat stopped “to wood”
every man servant rushed to the woodman’s cabin to get eggs,
chickens, milk, what not.
And those men had the privilege of the kitchen to prepare private
dishes for their white folks. I wonder how long a boat or hotel would
stand that kind of management to-day; but in the days where my
rocking-chair is transporting me, steamboat fare was not up to the
standard of any self-respecting pater familias. There was no ice
chest, no cold storage; in a word, no way of preserving fresh foods
for any length of time, so passengers resorted to such means as
presented themselves for their own bodily comfort. Those who had
not the necessary appendage—a man servant—foraged for
themselves, but the experienced and trusted servant, to use a
vulgarism, “was never left.”
The table for meals extended the length of the gentlemen’s cabin,
stretched out and out to its utmost length, if need be, so that every
passenger had a seat. There was no second table, no second-class
passengers—anybody was the equal of anybody else. If you could
not possibly be that, you could find accommodation on the lower
deck and eat from a tin plate.
It was quite customary, as I have mentioned, for passengers to
have private dishes, prepared by their own servants. I recall with a
smile, on one occasion, a very respectable-looking stranger boarded
our boat at Helena or some such place. At dinner he reached for a
bottle of wine. Cuthbert Bullitt touched the bottle with a fork, saying,
“Private wine.” The man, with a bow, withdrew his hand. Presently he
reached for a dish of eggs. My father said, “Excuse me, private.”
There was something else he reached for, I forget what, and another
fellow-passenger touched the dish and said “Private.” Presently
dessert was served, and a fine, large pie happened to be placed in
front of the Helena man. He promptly stuck his fork into it. “By
gracious! this is a private pie.” There was a roar of laughter.
After dinner the others, finding him delightfully congenial and
entertaining, fraternized with him to the extent of a few games at
cards. He was wonderfully lucky. He left the boat at an obscure river
town during the night, and the next day our captain said he was a
notorious gambler. From his capers at table the captain saw he was
planning a way of winning attention to himself, therefore under cover
of darkness he had been put ashore. My father, who did not play,
was vastly amused when he found the smart gambler had carried off
all the spare cash of those who had enjoyed the innocent sport.
Flatboats floating all manner of freight down stream were a
common sight on the river. Arrived at their destination, the boats,
which were only huge rafts with no propelling power, were broken up
and sold for lumber, and the boatmen traveled back up stream in
packets to repeat the process. Cousin Eliza Patrick used to relate
the trip her family made in about 1820 on a flatboat from Kentucky to
Louisiana. The widowed mother wished to rejoin a son practicing
medicine in the latter state, so she sold her land, and loaded her
family and every movable object she possessed—slaves, cattle,
farm implements, household effects—upon a huge “flat” and they
floated by day and tied up to the bank by night, carrying on, during