Load balancing in cloud computing

Prof. Dr. Noman Islam

Post-doc Fellow, University of Kuala Lumpur, Malaysia

Professor, Iqra University, Pakistan

International E-Conference on Recent Advances in Computer Science and Information Technology (RACSIT2020)

​

September 24-25, 2020

Keynote Speech

What is cloud computing?

• Cloud computing is defined as computing as a utility provided remotely over the internet via dynamic provisioning of services
• Service models [5]:
• Infrastructure as a Service (IaaS)
• Platform as a Service (PaaS)
• Software as a Service (SaaS)
• Deployment models [5]
• Public cloud
• Private cloud
• Hybrid cloud
• Community cloud

Challenges of cloud computing

• Some of the challenges of cloud computing includes [5, 6]:
• Protection of data
• Data retrieval and availability
• Administrative capabilities
• Regulatory and compliance restrictions
• Security
• Load balancing
• Fault tolerance
• Interoperability
• Portability

​

Objectives of research

1. To identify major challenges pertaining to load balancing in cloud computing
2. To analyze currently available load balancing strategies and propose a taxonomy
3. To identify load balancing metrics
4. To identify future research directions for load balancing
5. To propose a novel load balancing solution based on Long Short Term Memory (LSTM)

Search criteria

• We performed a systematic literature review of load balancing research papers over past five years
• The keywords used for search were cloud computing, load balancing, fault tolerance, machine learning, reinforcement learning, load balancing survey
• The papers published by popular publishers such as Springer, Inderscience, IEEE, ACM, Elsevier, Taylor & Francis were consider
• Further selection criteria was papers who are published in journals and specifically targeted load balancing issue of cloud computing
• Detailed results are published in [6]

Load balancing in cloud computing

• The objective of load balancing is to distribute job requests amongst the virtual machines to ensure that hosts are neither under-loaded nor overloaded
• The load balancer distributes the loads to virtual machines optimally to maximize the utilization of resources
• NP-complete problem
• Types of load balancer [1]:
• Hardware
• Network
• Application

Load balancing in cloud computing [1]

Load balancing strategies [1, 2]

• Threshold based

​

• Static Vs Dynamic

​

• Sender-initiated Vs receiver initiated Vs symmetric

• Centralized Vs decentralized

​

• Offline Vs batch-mode

​

• A number of strategies have been proposed in literature [1,2,3,4]

​

A taxonomy of load balancing strategies

• Load balancing
• Classical
• Round Robin

​

• First Come First Served (FCFS)

​

• Minimum Execution Time (MET)

​

• Threshold-based

​

​

• General approaches
• Workflow based

​

• Map-reduce based

​

• Divide and conquer

​

• Max-min

​

​

• AI-based
• Optimization
• Swarm intelligence
• Ant colony

​

• Honey bee inspired

​

• Particle swarm optimization

​

• Dragonfly optimization

​

• Gray-wolf optimization

​

​

• Operation research

​

​

• Heuristic based

​

• ML based
• Supervised

​

• Reinforcement learning

​

​

• Game theory

​

​

Load balancing challenges [2, 3, 4]

• Geographical Distributed Nodes

​

• Single Point of Failure

​

• Virtual Machine Migration

• Heterogeneous Nodes

​

• Algorithm Complexity

​

• Virtual machine migration

​

Load balancing parameters [1, 3, 4]

• Throughput
• Thrashing
• Reliability
• Accuracy
• Makespan
• Scalability

• Fault-tolerance
• Migration time
• Energy consumption
• Response time
• Degree of imbalance
• Carbon emission

​

An LSTM based load balancing solution

• This work proposes an LSTM based solution for load balancing in cloud environment
• It is hypothesized that the virtual machine load can be characterized as a sequential data and LSTM can be used to predict future load [7]
• Based on predicted load, jobs can be scheduled

Long Short Term Memory model (LSTM)

• Long Short Term Memory model (LSTM) belongs to the family of recurrent neural network (RNN)
• RNN has long been used for modeling sequential data
• In sequential data, the future state depends on current state
• However, RNN is not good at modeling long-term dependencies
• Suffers from vanishing gradient and exploding gradient problem
• The solution is LSTM that uses various gates to model these dependencies

LSTM-based Load Balancer

​

H1

H2

H3

H4

Model

Matcher

Jobs

Hosts

An LSTM-based load balancing approach for cloud computing

Future directions

• ML is emerging as a technique for load balancing
• Datasets are needed
• What happened in case of false positive
• Thrashing issue in VM migration
• Security consideration before VM migration [3]
• Energy efficient load balancing [3, 4]

References

[1] S. K. Mishra, B. Sahoo, and P. P. Parida, ‘‘Load balancing in cloud computing: A big picture,’’ J. King Saud Univ.-Comput. Inf. Sci., vol. 32, no. 2, pp. 149–158, Feb. 2020

[2] P. Kumar and R. Kumar, ‘‘Issues and challenges of load balancing techniques in cloud computing: A survey,’’ ACM Comput. Surveys, vol. 51, no. 6, pp. 1–35, Feb. 2019

[3] MOHAMMED ALA’ANZY, MOHAMED OTHMAN, “Load Balancing and Server Consolidation in Cloud Computing Environments: A Meta-Study”, IEEE Access, 2019

[4] E. J. Ghomi, A. M. Rahmani, and N. N. Qader, ‘‘Load-balancing algorithms in cloud computing: A survey,’’ J. Netw. Comput. Appl., vol. 88, pp. 50–71, Jun. 2017

[5] N. Islam, A U Rehman, “A comparative study of major service providers for cloud computing”, in proceedings of 1st International Conference on Information and Communication Technology Trends, 2013, Karachi, Pakistan

[6] MA Shahid, N Islam, M Alam, MS Mazliham, S Musa, “A comprehensive study of load balancing approaches in the cloud computing environment and a novel fault tolerance approach”, IEEE Access, 2020

[7] Yonghua Zhu, Weilin Zhang, Yihai Chen & Honghao Gao, “A novel approach to workload prediction using attention-based LSTM encoder-decoder network in cloud environment”, EURASIP Journal on Wireless Communications and Networking , 2019