Table 1. Cloud Provisioning
Mobile networks have gone through various stages of evolution with each stage aimed at addressing a wide range of challenges and limitations. During the early evolutions of mobile networks 2G, 2.5G, 3G are the key challenge to investigate in efficient and cost effective ways to delivering high data speeds. This has led to the proposals and development of 4G LTE networks based on a flat all-IP architecture with Internet based protocols. However, recent trends indicate that the Internet-like architecture on mobile networks has further enabled Internet based cloud service providers to provide Over- The-Top (OTT) applications to mobile devices bypassing, and competing with the mobile operator on services, such as voice, video, messaging, and gaming. This is a key motivation for cloud service providers and mobile operators to explore various opportunities in which they can both leverage on their existing infrastructures in order to efficiently deploy cloud services in mobile environments. In this paper, we study the challenges and limitations that constrains the efficient deployment of cloud services in mobile environments. Then we propose a collaborative approach in which the cloud service provider, and mobile operator can dynamically manage the underlying mobile network infrastructure resources to optimise the delivery of cloud services in mobile environments. This is achieved by using cloud management approaches with the ability to factor such as mobility and frequency spectrum of mobile resources while integrating the cloud service provider, and mobile operator cloud infrastructures.
Over the recent years, the wide adoption of virtualization technologies, ubiquitous broadband access, and the exponential demand for Internet based services have led to the emergence of cloud services models that enables the delivery of infrastructure, platform or application services from large scale hosted infrastructures using highly efficient and cost effective approaches (Rohith & Anand, 2018). The paradigm of cloud services is not limited to fixed devices, but the exponential rise of mobile users who assessing cloud services via Internet enabled smart mobile devices is currently driving cloud service providers and mobile network operators to address the current limitations of deploying cloud service in mobile environments. This is because mobile devices utilize the resource from cloud providers without complex hardware and software implementations at the device side. However, due to the mobility of mobile users, location based (or geo-based) cloud resource provisioning is required to reduce the end-to-end communication delay (Deepa & Cheelu, 2017). With the emergence of high-speed Long Term Evolution (LTE) networks, more mobile devices will connect natively to mobile networks seamlessly integrating LTE and W-iFi technologies. This explosion of mobile devices including notebooks, mobile phones, tablets, sensors, and televisions to mention a few, can cause a wide range of constraints and limitations on mobile networks. However, if these limitations and constraints are not addressed rapidly, they may lead to the degradation of cloud services, while negatively affecting the Quality of Experience (QoE).
A collaborative approach is proposed where the cloud service provider and mobile operator can dynamically manage the basic mobile network infrastructure resources in order to improve the delivery of cloud services in mobile environments. This is achieved by using cloud management approaches with factors in mobile resources such as mobility and frequency spectrum while integrating the cloud service provider and mobile operator cloud infrastructures (Rai & Anand, 2017).
Mobile networks have gone through various stages of evolution, each stage aimed at addressing a wide range of challenges and limitations. Mobile networks such as 2G, 2.5G, 3G and the key challenges of delivering high data speeds. This has led to the development of 4G LTE networks based on a flat all-IP architecture with Internet based protocols. Recent trends indicates that the Internet-based cloud service providers are offering Over- The-Top (OTT) applications to mobile devices, bypassing and competing with the mobile operator on services such as voice, video, messaging etc. However, the mobile operator still has full control of mobile network and can influence the Quality of Experience (QoE) for the mobile user. This is an important motivation for cloud service providers and mobile operators to explore various opportunities for efficient delivery of cloud services in mobile environments. To help further demonstrate the current challenges faced by this collaboration, we have presented three use case scenarios in which independent solutions can be provided by either Internet cloud service providers or mobile operators. However, the most efficient and cost effective solutions will be provided by a collaboration in which both parties leverage on their existing solutions (Kumar & Anand, 2017).
1.1.1 Smart Metering
In this scenario, an utility service provider has deployed smart meters to its end-users premises in order to provide meter readings back to its cloud application hosted on the Internet. Primarily, data is sent via the end- user's W-iFi connection, but the smart meters are also configured with backup SIM cards as a redundant path through the mobile network in case the W-iFi network is no longer available. However, the utility provider demands for an end-to-end secured solution so as to have control of the data as it tranverse through the network (Shoja, Nahid, & Azizi, 2014).
1.1.2 Crop Imaging
In this scenario, a local farmer discovers strange patterns in the crops growing on his farm. In order to get more details on the patterns, he takes a picture of the pattern on his mobile phone, and uploads the picture and his location to the cloud service. The service analyses the images against well known crop diseases by matching over 3 million images, starting with locally reported crop diseases before expanding the search globally. This whole process takes five minutes, as it is computing intensive. In this, the cloud service provider provision its infrastructure to support the farmer's search, whether or not the farmer reports any crop diseases. On another note, if there is an outbreak of a crop disease in a local area, the cloud service provider will need to dynamically provide local on-demand computing resources for all local request as close to the farmer as possible (Arvindhan, Anand, & Viswanathan, 2019).
1.1.3 Community Services
In this scenario, users living and working within the same local community set up an online service that provides content specifically for its citizens. An example of such as service is a University based television network targetted towards students on its campus. Primarily, students will connect to these community services via the University's local W-iFi networks, but as soon as they are out of range of any W-iFi networks, streams are dynamically diverted via mobile operators network and routed via are Internet back to the University servers. As these streams contribute to a large amount of traffic on the mobile operator's network and the University internet router's, the University and mobile operator can potentially investigate various approaches in which the traffic can dynamically localised with minimal impact on resources (Liang, Cai, Huang, Shen, & Peng, 2012).
Based on the scenarios explored above, it is important to note that the key challenge highlighted in these scenarios is the ability to enable cloud resources to be dynamically managed across the Internet and mobile networks. In consideration of the suitability of current management capabilities applied as cloud solutions, it is necessary to appreciate those components, which need to be managed.
Core elements of a cloud include its storage capacity, communication ability (network provision), compute capability (typically servers) and other hardware devices, such as switches and routers. Management of such components must be dynamic due to the way in which such resources are rolled out. Furthermore, applications may accommodate variation in the service level provided by the network in poor operating conditions. Resource availability changes not only in response to application demand, but also due to virtual resource provision. Virtualisation ability allows the residual capacity to achieve Quality of Service (QoS) by responding to specific application demand, optimise resource consumption across the cloud by provisioning only those resources demanded (with a small degree of resource redundancy), and achieving it efficiently (Anand, Sihag & Gupta, 2012). Virtual servers, storage, and network capacity are adapted dynamically during cloud management in terms of their existence, operational state (i.e., active, idle, sleeping), location and replication across clouds, as a range of examples. Careful management of cloud resources is important to accommodate the efficient achievement of application Service Level Agreements (SLAs). Network traffic to from and within a cloud includes management, backup, storage, and service traffic. Management traffic includes that, which collects real-time context on the network and distributes information to the management agent for evaluation and subsequent enforcement of informed decisions. Careful management of cloud resource availability is important to accommodate the information flow (Kumar & Anand, 2017).
Resources require management across clouds due to their dynamic availability, overhead associated with management and cost of adapting resources based on decisions applied by the management system. Resource management, which is responsive to network activity, is important due to service differentiation across cloud providers and a desire to maximise financial return through optimising service provision. Lack of standardised approach, availability of storage and service resources from different providers, cloud management by different entities and support of applications across multiple clouds further complicates the interoperability processes when provisioning an end-to-end networking solution. This is particularly important in clouds, where resource availability can change rapidly when users have variable application requirements, and management overhead adapts in response to client interactivity with the cloud. Such characteristics occur in response to the suitability of monitoring procedures, which influence the actions enforced in terms of resource provisioning, adaptive monitoring, and resource replication.
In this section, the state of art in cloud management is reviewed to allow requirements for interoperability, together with awareness of solution limitations, to be appreciated.
In this section, a high level overview of selected cloud management platforms are provided to reflect aspects of the different cloud management functions:
The resource provisioning function within cloud management technology aims to automate the rapid abstraction, deployment, customisation, and retirement of virtual resources using Application Programmable Interfaces (APIs) or a Graphical User Interface (GUI). This involves the use of efficient algorithms and technologies to carry out various tasks, such as resource scheduling and allocation, fault tolerance, work load distribution, auto scaling and energy management.
As shown in Table 1, a number of cloud management systems already provide resource provisioning capabilities within their product suite. However, management applications such as RightScale, Scalr, and enStratus have been developed specifically to manage a wide range of cloud platforms and focus on more supporting generic capabilities across a wide range of environments. This is commonly achieved by communicating with the compute, storage and network controllers based on overall cloud management goals.
Table 1. Cloud Provisioning
With regard to the compute controller, cloud provisioning is commonly achieved using a hypervisor, as described in Section 1.1.2. However, Table 1 shows the most supported hypervisor among the selected cloud management platforms is Xen, followed by KVM, and the least supported being Microsoft's Hyper-V. Although Xen does not have any particular technical advantages over KVM with regard to cloud provisioning, it may be more widely supported because it has been available longer than KVM, with effort invested in extending and customising it for a range of architectures and platforms. One disadvantage of Xen is that there are many variations due to different customisation needs. Some examples include Citrix XenServer, Oracle VM, Amazon Xen and Open VM. This makes it more difficult to maintain consistency and standardisation across all Xen releases. However, as KVM is relatively new, there has been a push to standardise all processes and interfaces across supported platforms. Furthermore, in order to drive standardization of KVM, Xen has been dropped as the default hypervisor for most Linux distributions and replaced with KVM. Currently, most management platforms support two hypervisors, but with a greater drive towards Xen, as it is only hypervisor support on the most influential cloud platform Amazon Web Services.
As shown in Table 2, cloud interoperability is observed across the major cloud providers: VMware, For example, joined the OpenStack Foundation in September 2012, an action which enables greater networking capability between resources provisioned by each individual platform. CloudStack has been released to Apache to support open source cloud hosting. For example, Vmware can be supported in CloudStack since February 2012, while CloudStack and OpenStack on the other hand, are often seen to be competitors of each other. CloudStack is perceived, for example, to have a greater level of maturity then OpenStack.
Table 2. Cloud Interoperability
RightScale and Eucalyptus are another partnership in which the qualities of each may be exploited in myCloud. RightScale supports roll out of interoperable hybrid clouds using the public cloud, with Eucalyptus enabling fast setup of a private cloud. A range of cloud services are available using RightScale and Eucalyptus together, including myCloud Free, myCloud Standard, and myCloud Enterprise.
Based on the limitations highlighted in Section 1 and 2, we propose a function that provides a dynamic and programmable interface easily integrates mobile and public Internet domains to allow for the deployment of cloud based services in mobile environments. In this section, we provide high level requirements and specifications that need to be considered in the design of such a function.
A key requirement is the ability to identify the resource footprint of cloud based applications, due to their competing resource requirements. Furthermore, another requirement is to identify user behaviour and explore its predictability. Together, these allow estimations of service demand and resource requirements that may be used to influence the proposed design of such a function. This is important because the function must have the ability to carrying out cloud management functionalities so as to make key decisions, such as the placement of data, replication of data, placement of management proxy, size of the cloud, and characteristics of server resources deployed (such as disk, memory and CPU size) need to the cloud service to be functional in cloud environment. In simpler terms, the function will serve as a broker on behalf of remote resources, which are not part of a centralised data center or mobile network, with the function of offering, part or all of the needed resources i.e. compute, storage and network resources to an application.
Traffic from the mobile environment aimed at cloud services on the Internet is intercepted by the function in order to dynamically provision the resources needed for the particular cloud service session. Although there exist well- established interfaces and protocols to integrate the mobile and fixed environments, this function will provide an aggregation point that will allow the dynamic and programmatic invocation of these interfaces and protocols. Furthermore, as there already exists a wide range of cloud service providers with varying contractual arrangements with different Mobile Network Operators (MNO), this cloud management function will allow for mobile network operators to federate all managed cloud based services aimed for their mobile network, while abiding to varying needs and SLAs of different cloud service providers.
Integration of mobile networks to the public Internet is still one of the main driving forces toward the evolution of the current Internet architecture. This is because it plays an important role in the “Future Internet” theme aimed at integration of our physical and digital worlds via the use of mobile sensing devices in form of mobile phones, sensors, and actuators. This will open up new services and opportunities in the area of Internet of Things (IoT) and Machine-to-Machine (M2M) communications driving innovative visions, such as smart cities, personal healthcare and e-agriculture.
In this study, we have presented the challenges and limitations that limiting the effective use of cloud services in mobile environments. In order to improve the delivery of cloud services in mobile environments, we recommend a joint operation of the cloud service provider and mobile operatior to dynamically manage the basic mobile network infrastructure resources. This is achieved by using a cloud management approaches with the ability to factor mobile resources such as mobility and frequency spectrums while integrating the cloud service provider and mobile operator cloud infrastructures.