Keywords

1 Introduction

Cloud computing has recently emerged as one of the most promising and challenging technologies. It is based on a computing paradigm where a large pool of systems are connected in private, public or hybrid networks, to provide dynamically scalable infrastructure for computing resources [19]. The computing resources are available to the users via the internet [13]. The characteristics of cloud computing include on-demand self service, broad network access, resource pooling, rapid elasticity and measured service. On-demand self service means that organizations can access and manage their own computing resources. Broad network access allows services to be offered over the Internet or private networks. Pooled resources mean that customers draw from a pool of computing resources. Services can be scaled larger or smaller; and use of a service is measured. The cloud computing service models are Software as a Service (SaaS), Platform as a Service (PaaS) and Infrastructure as a Service (IaaS) [10]. In Software as a Service model, consumer uses the provider’s applications running on a cloud infrastructure. Example of SaaS is Salesforce [2]. In PaaS, an operating system, hardware, and network are provided, and the customer installs or develops its own software and applications. The most prominent key players of PaaS are Azure platform [8] and Google App Engine [5]. The IaaS model provides just the hardware and network; the customer installs or develops its own operating systems, software and applications. The examples of IaaS are Amazon EC2 service [1], GoGrid [4], Flexiscale [3], and Redplaid [7].

Cloud services are deployed as a private cloud, community cloud, public cloud or hybrid cloud [10]. In public cloud, services are offered over the Internet and are owned and operated by a cloud provider. In a private cloud, the cloud infrastructure is operated solely for a single organization, and is managed by the organization or a third party. In a community cloud, the service is shared by several organizations and made available only to those groups. The infrastructure may be owned and operated by the organizations or by a cloud service provider. A hybrid cloud is a combination of two or more cloud infrastructures (private, community, or public) that remain unique entities, but are bound together by standardized or proprietary technology that enables data and application portability. Virtualization technologies are usually used to access computing resource by the users. Users can specify required software stack such as operating systems, software libraries, and applications, then package them all together into virtual machines (VMs). VMs will be hosted in cloud service providers. Lightweight cloud computing infrastructures combine cloud and grid computing concept to provide a shared infrastructure over commodity hardware such as mobile phones, desktop, tablets, etc. As the cloud computing market expands and the number of users and cloud service providers increases, there is a need for a centralised system [22], called cloud services brokerage (CSB), to optimize resource allocation by managing and monitoring the activities of cloud users and cloud service providers. For the CSB to work efficiently, a reliable database management system needs to be implemented on the CSB site to keep and update the track of customer requests and cloud infrastructures status. Relational database management systems (RDBMS) otherwise known as SQL database cannot cope with the unprecedented scale factors that modern cloud-based applications have introduced. The cloud applications need to support large numbers of concurrent users and be able to handle unstructured and semi-structured data. To solve this problem, NoSQL (Not Only SQL) databases emerge to support large-scale application demands. In addition, previous studies have shown that NoSQL databases perform better than SQL databases especially in the cloud computing environment where there are occurrence of huge volume of data [27].

1.1 Contributions and Outline

In this paper, we present a data model for cloud computing environment to help the CSB support the allocation, control and management of virtual resources between CSPs and cloud users. We implement this model using the graph database (Neo4j) and document-oriented database (mongodb) on our private lightweight cloud testbed, using a syntactic of cypher language to store, update and retrieve the customer requests and cloud infrastructures status in the database. Building upon the free and open source OpenStack software platform for cloud computing, the model is intended to provide infrastructure-as-a-service (IaaS) in community sensor networks [29] for applications such as drought mitigation for small scale farming [30, 31] and cyber healthcare [32, 33] in the rural areas of the developing countries. Potential applications which might also benefit from this model include smart parking [34], pollution monitoring [35] and public safety [36] in the smart developing cities. The rest of this paper is organized as follows; In Sect. 2, we describe the concept of cloud service brokerage system. The next section presents previous studies related to the management of virtual resources information in data center and cloud computing. Section 4 describes proposed cloud computing environment model. The data model of cloud computing environment is presented in Sect. 5. Implementation of the models and experimental results are found in Sects. 6 and 7, and finally, we conclude the paper in Sect. 8.

Fig. 1.
figure 1

Cloud computing environment

Full size image

2 Cloud Services Brokerage

As depicted in Fig. 1, the cloud computing environment considered in this paper consists of user, virtual machine repository (VMR), and cloud services broker and CSP, which consists of physical machines (PM) and data center (DC). A cloud services broker is a third-party individual or business acting as a middle man between cloud service users and CSPs. Cloud service brokers rent different types of cloud resources from many cloud Service providers and sublet these resources to the requesting cloud users. The cloud service broker performs the following functions: (i) optimal placement of the virtual resource of a virtual infrastructure across multiple cloud service providers; (ii) management and monitoring of these virtual resources; and (iii) aggregation of multiple cloud services into one or more customer-tailored cloud services. OPTIMIS [6] identifies the requirement and capabilities that a cloud service broker needs to have in order to play the role of brokerage services:

  • Effectively match the requirements of the cloud user with the service provided by the CSPs.

  • Negotiate with CSPs and cloud users over service level agreements (SLA).

  • Effectively deploy services of CSP onto the cloud users.

  • Maintain performance check on these SLA’s and take actions against SLA violations.

  • Ensure data confidentiality and integrity of CSP’s service.

  • Enforce access control decisions uniformly across multiple CSPs.

  • Securely map identity and access management systems of the CSPs.

However, an effective database management system needs to be included as one of the functional components of a CSB.

3 Related Work

This section presents two broad categories of related work. The first category discusses existing cloud brokerage systems and the second category presents the related work to the database model in cloud computing.

3.1 Existing Cloud Brokerage Systems

Many broker-based systems have been proposed to solve cloud computing problems. Heilig et al. [16] propose a cloud brokerage approach to solve the Cloud Resource Management Problem in multi-cloud environments with aim to reduce the monetary cost and the execution time of consumer applications using Infrastructure as a Service of multiple cloud providers. In [15], the authors propose a broker-based architecture and algorithm for placing and migrating virtual resources to physical machines. In [20], the authors propose a federated cloud computing environment in which a cloud broker has the ability to interface more than one cloud provider to support several users. These Users access cloud services via web interface. The cloud service broker pays the usage of the cloud resources to the cloud service provider, and charges the user for these resources. In [25], a solution to manage the information of a large number of cloud service providers via a unique indexing technique is proposed. STRATOS [21] proposes a cloud brokerage service that solves a Resource Acquisition Decision (RAD) problem in the selection of n resources from m cloud services. In [18] develops a cloud brokerage service for measuring the performance of a range of cloud services including; elastic compute clusters, persistent storage, intra-cloud networking and wide-area networking. [17] proposes a novel secure sharing mechanism for a secure cloud bursting and aggregation operation in which the cloud resources are shared in a confidential manner among different cloud environments.

3.2 Database Model in Cloud Computing

Goli-Malekabadi et al. [14] proposes an effective database model for storing and retrieving big health data in cloud computing. The study presents the model based on NoSQL databases for the storage of healthcare data and was implemented in the cloud environment for gaining access to the distribution properties. The experimental results of the model was evaluated with relational database model in terms of query execution time, data preparation, flexibility and extensibility parameter. The results show that the proposed model outperforms the relational database. In [28], the authors propose a novel protocol to enable secure and efficient database outsourcing. First, the authors propose a new cloud database model by introducing computation service providers which can accommodate the conventional DBaaS model and introduce a proposed database outsourcing protocol secureDBS which uses a secret sharing mechanism. The experiments conducted show that the proposed model is reliable, secure and efficient. In [12], the authors propose a novel management scheme that enables the representation and the retrieval of (structured or unstructured) big data using conceptual graphs and structured marks. Curino et al. [11] proposes relational database as-a-service for the cloud. This work describes the challenges and requirements of a large-scale, multi-node DBaaS and presents the design principles and implementation status of relational cloud. The advantage of this work is that it addresses three significant challenges, which are: (i) efficient multi-tenancy; (ii) elastic scalability; and (iii) database privacy.

However, none of these works have proposed a graph data model and its implementation using graph database as of the requirement for the effectively cloud brokerage services.

Fig. 2.
figure 2

Graph model for cloud computing environment

Full size image

4 Cloud Computing Environment Model

In this section, we introduce the system model for our cloud computing environment. As depicted in Fig. 1, the cloud computing environment consists of User, Virtual Machine Repository (VMR), Cloud Service Provider (CSP), Physical Machines (PM), Data Center (DC) and Cloud Services Broker (CSB).

We consider a set CSP,

$$\begin{aligned} CSP = \lbrace csp_{1}, csp_{2}....csp_{n} \rbrace \end{aligned}$$
(1)

where n is the number of CSPs managed by the CSB. Each CSP consists of DC,

$$\begin{aligned} DC = \lbrace dc_{1}, dc_{2}....dc_{m} \rbrace \end{aligned}$$
(2)

where m is the number of DCs in a CSP and each DC contains a member of PMs,

$$\begin{aligned} PM = \lbrace pm_{1}, pm_{2}....pm_{q} \rbrace \end{aligned}$$
(3)

where q is the number of PM in a DC and each PM hosts t number of VMs as expressed by the set

$$\begin{aligned} VM = \lbrace vm_{1}, vm_{2}....vm_{t} \rbrace \end{aligned}$$
(4)

We consider that K jobs need to be allocated to CSPs. Each job k requires \(w_{j}\) number of virtual machines. The CSB allocation is expressed by the notation

$$\begin{aligned} k \longrightarrow vm_{w_{j}} \end{aligned}$$
(5)

subject to \(w_{j}\) = 1 or \(w_{j}\) \(\le \) t. Each vm is placed in one pm,

$$\begin{aligned} vm \longrightarrow pm \end{aligned}$$
(6)

Each pm is hosted by one data center dc,

$$\begin{aligned} pm \longrightarrow dc \end{aligned}$$
(7)

Finally, each data center is owned by one cloud service provider (csp),

$$\begin{aligned} dc \longrightarrow csp \end{aligned}$$
(8)

5 Data Model of Cloud Computing Environment

There are many different NoSQL data models and each one of them has a different structure. In this section, we present the deployment of graph and document-oriented models for cloud computing environments.

5.1 Graph Model

Graph data models have emerged with the objective of modeling information whose structure is a graph [9]. It encodes entities and relationships between entities using directed graph structure [23]. It is a set of vertices and edges where vertices denote nodes and edges represent relationship between these Nodes. Graphs are data structures for storing data that is heterogeneously structured. Graphs can be directed or undirected. Undirected graphs can be traversed in both directions while directed graphs can be traversed only in one direction. The properties of a graph model includes the following [24];

  • It contains nodes and relationships.

  • Nodes contain properties (key-value pairs).

  • Nodes can be labelled with one or more labels.

  • Relationships are named and directed, and always have a start and end node.

  • Relationships can also contain properties.

The graph model of cloud computing environment can be represented mathematically as a graph G(V, E) where V is the set of resource nodes and E is the set of relationships between the nodes such that,

$$\begin{aligned} \lbrace CSP, DC, PM, VM, K \rbrace \subset V \end{aligned}$$
(9)

Constrained by the notation:

$$\begin{aligned} |DC| \ge |CSP| \end{aligned}$$
(10)
$$\begin{aligned} |VM| \ge |PM| \end{aligned}$$
(11)
$$\begin{aligned} |K| \ge |VM| \end{aligned}$$
(12)

The relationships between the nodes of the graph (V, E) are defined as follows:

  • \(csp \rightarrow dc\) represents the relationship between cloud service provider and data center.

  • \(dc \rightarrow pm\) represents the relationship between data center and physical machine.

  • \(pm \rightarrow vm\) represents the relationship between physical machine and virtual machine.

  • \(vm \rightarrow k\) represents the relationship between virtual machine and job.

Such that,

$$\begin{aligned} csp \rightarrow dc, dc \rightarrow pm, pm \rightarrow vm, vm \rightarrow k \subset E \end{aligned}$$
(13)

The graph is illustrated by Fig. 2.

5.2 Document-Oriented Database

In a document-oriented model, data objects are stored as documents; each document stores data which can be updated or deleted. Instead of columns with names and data types, data is described in the document, and provide the value for that description. The difference between a relational model and a document-oriented model is as follows: in a relational model, data is added by modifying the database schema to include the additional columns and their data types while in document-based data, additional key-value pairs will be added into documents to represent the new fields. The document-oriented model for cloud computing environment is represented in Fig. 3.

Fig. 3.
figure 3

Document-oriented model for cloud computing environment

Full size image

6 Experiments

We conducted two different experiment in order to evaluate the both graph and document-oriented models on a proxy node of our private lightweight cloud testbed running on Openstack architecture. The proxy node is a Linux Machine with an Inter(R) core(TM) i5-4590, 3.30 Ghz CPU, 8 GB RAM and serves as a cloud brokerage system which initiate upload and download operations across multiple storage nodes. The CSB can update the cloud resources status in database either manually or through real-time process by building Application Protocol Interface (API) connecting cloud infrastructures to a database. We explore the suitability of two different databases in a cloud environment, the database are: (i) graph database: Neo4j; and document-oriented database: MongoDB respectively.

6.1 Implementation of Graph Database

Graph databases are databases that support graph model. One of the examples of graph database is Neo4j and it can be accessed using cypher query language. The graph database is implemented using Neo4j Community version 3.0.1. The database contains nodes with labels, properties and relationship between them as follows:

  • csp(‘CLOUDPROVIDER’, cspid, name, cost, costPerMem, costPerStorage, costPerBw)

  • dc(‘DATACENTER’, name, centerid, location, arch, os, time_zone)

  • pm(‘PHYSICALMACHINE’, name, pmid, mips, ram, bw, storage)

  • vm(‘VIRTUALMACHINE’, name, vmid, mips, ram)

  • job(‘JOB’, job_id, name, length, filesize).

6.2 Cypher Query Language

Here, we discuss the Cypher syntax to create and retrieve nodes and relationships in graph database.

  • Cypher syntax to create Nodes and relationship.

    • Create Datacenter, Cloud provider nodes and relationship between them.

      CREATE (dc1:DATACENTER\(\lbrace \) name =‘dc1’, centerid =1, location =‘Capetown’, arch = “x86”, os =“Linux”, time-zone = 10.0 \(\rbrace \) ) -[:OWNED-BY]-> (csp:CLOUDPROVIDER \(\lbrace \) name = ‘csp1’, cspid = 0, cost = 3.0, costPerMem = 0.05, costPerStorage = 0.001, costPerBw = 0.0 \(\rbrace \))

    • Create Datacenter, Physical Machine nodes and relationship between them.

      CREATE (pm:PHYSICALMACHINE\(\lbrace \) name =‘pm1’, pmid = 0, mips = 1000, ram = 20148, bw = 1000, storage = 1000000 \(\rbrace \) ) -[:DEPLOYED-IN]-> dc1:DATACENTER \(\lbrace \) name = ‘dc1’, centerid = 1, location = ‘Capetown’, arch = “x86”, os = “Linux”, time-zone = 10.0 \(\rbrace \))

    • Create Virtual Machine, Physical Machine nodes and relationship between them.

      CREATE (vm:VIRTUALMACHINE\(\lbrace \) name = ‘vm1’, vmid = 0, mips = 250, ram = 20148 \(\rbrace \) -[:HOSTED-BY]-> dc1:DATACENTER \(\lbrace \) name = ‘dc1’, centerid = 1, location = ‘Capetown’, arch = “x86”, os =“Linux”, time-zone = 10.0 \(\rbrace \))

    • Create Virtual Machine, Job nodes and relationship between them.

      CREATE (job:JOB\(\lbrace \) app_id = 0, name = ‘job1’, length = 4000, filesize = 24 -[:RUNS-ON]-> vm:VIRTUALMACHINE \(\lbrace \) name = ‘vm1’, vmid = 0, mips = 250, ram = 20148 \(\rbrace \)).

  • Cypher queries to retrieve information from graph database.

    • Query 1: Find all Virtual Machines assigned to Jobs.

      MATCH (a)-[:RUNS_ON]->(b) RETURN a.name as JOB, b.name as VIRTUALMACHINE;

    • Query 2: Find name of resources owned by Cloud Service Providers.

      MATCH (b)-[:HOSTED_BY]->

      (c)-[:DEPLOYED_IN]-> (d)-[:OWNED_BY]->

      (e:CLOUDPROVIDER) RETURN b.name as VIRTUALMACHINE, c.name as PHYSICALMACHINE,

      d.name as DATACENTER, e.name as CLOUDPROVIDER;

    • Query 3: Find all resources used by job 1.

      MATCH (a:JOB\(\lbrace \)name:‘job1’\(\rbrace \))-[:RUNS_ON]->(b)-[:HOSTED_BY]->(c) -[:DEPLOYED_IN]->(d)-[:OWNED_BY]->(e) RETURN a as JOB, b as VIRTUALMACHINE, c as PHYSICALMACHINE, d as DATACENTER, e as CLOUDPROVIDER;

6.3 Document-Oriented Database

Document-oriented databases are one of the NoSQL databases. A document-oriented database is designed for storing, retrieving, and managing document-oriented, or semi structured data. The main concept of a document-oriented database is the notion of a Document. MongoDB, CouchDB and Terrastore are examples of the Document-oriented databases. In this work, document-oriented database is implemented using MongoDB shell version: 2.4.9. Mongodb is very famous NoSQL databases in the data industry [26]. All the formats are loaded in JSON format. In MongoDB, data is grouped into sets that are called collections. Each collection has a unique name in the database, and can contain an unlimited number of documents. Collections are similar to tables in a relational database, except that they do not have any defined schema.

MongoDB Query Language. The Queries to create and retrieve collections in MongoDB re discussed below.

  • Commands to create Collections. In MongoDB, there is no need to create collection. MongoDB creates collection automatically, when inserting document.

    • Create JOB collection

      db.JOB.insert( \(\lbrace \) job_id : 0, name : ‘app1’, length : 4000, file-size : 24, vm_id : 0 \(\rbrace \) )

    • Create Virtual Machine collection

      db.VIRTUALMACHINE.insert( \(\lbrace \) vm_id : 0, job_id : 0, name : ‘vmm1’, mips : 250, ram : 20148, pm_id : 0 \(\rbrace \) )

    • Create Physical Machine collection

      db.PHYSICALMACHINE.insert( \(\lbrace \) pm_id : 0, job_id : 0, name : ‘pm1’, mips : 1000, ram : 20148, storage : 10000, vm_id : 0, center_id : 0 \(\rbrace \) )

    • Create Data center collection

      db.DATACENTER.insert( \(\lbrace \) center_id : 0, pm_id : 0, name : ‘dc1’, OS : Linux, location : ‘Durban’, arch : ‘x86’, time_zone : 10, csp_id : 0 \(\rbrace \) )

    • Create Cloud Provider collection

      db.CLOUDPROVIDER.insert( csp_id : 0, center_id : 0, name : ‘csp1’, cost_per_BW : 0, cost_per_storage : 1, cost : 3 \(\rbrace \) )

  • Mongodb queries to retrieve information from relational database.

  • Query 1: Find all Virtual Machines assigned to Jobs.

    db.VIRTUALMACHINE.find( \(\lbrace \) JOB: \(\lbrace \) vm_id: [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23] \(\rbrace \) \(\rbrace \) )

  • Query 2: Find name of resources owned by Cloud Service Providers.

    figure a
    Table 1. Cloud computing entities
    Full size table

  • Query 3: Find all resources used by job 1.

    figure b
Table 2. Comparison of graph and document-oriented databases
Full size table

7 Experimental Results

We consider two cases as shown in Table 1. where the number of cloud infrastructures are varied. We run each query 5 times on each database for the two cases, execution time are recorded and the average of the execution times are calculated for each query. All times are measured in milliseconds. The result for the queries on the databases are presented in Table 2.

It can be observed from the values in Table 2 that the times taken to process queries for mongoDB are less when compared to that of Neo4j. For instance, in case 1 and query 2, mongoDB takes 5.6 ms while Neo4j takes 8.6 ms. Further analysis of the results show that the time taken to process case 1 and query 2 on MongoDB is less by 28% than that of Neo4j. It can also be deduced from the results that as the number of cloud computing elements increases, the query processing times gets increased manifoldly. More specifically, the time taken to process queries for case 2 is higher than that of case 1 due to the varied increased parameter in case 2 as opposed to lower parameter values in case 1.

8 Conclusion and Future Work

Cloud service brokerage system is a third party system that acts as a middleman between users and cloud service providers. However, for cloud service brokers to remain relevant in the cloud computing era, there is need to adopt an effective database model that can withstand the unprecedented demand from cloud users and providers. Hence in this research, we present a novel data model and explore the suitability of these database models in a cloud computing environment. The models are: (i) graph; and (ii) document-oriented models. We implement the models on our private cloud network testbed using Neo4j and MongoDB databases respectively. We also present query syntax to retrieve information from the databases. Finally, we compare the efficiency of the these database models in terms of query processing time, and varied the experimental parameters in order to establish the suitability of the models in a cloud computing environment. The experiment results show that document-oriented model has better performance in a cloud computing environment than graph modes, in terms of queries processing time. Ultimately, MongoDB emerges as the most suitable database model with respect to flexibility, elastic scalability, high performance, and availability [37,38,39].

In future, an optimization module can be developed on top of mongoDB database in cloud service brokerage system. The module will interface the system with cloud service providers and updating the status of cloud resources in the database.