Applying the Cloud Computing Model in LTE based
Cellular Systems
Arjan J. Staring
University of Twente P.O. Box 217, 7500AE Enschede
The Netherlands
[email protected]
ABSTRACT
As cloud computing emerges as the next novel concept in computer science, it becomes clear that the model applied in large data storage systems used to resolve issues com-ing forth from an increascom-ing demand, could also be used to resolve the very high bandwidth requirements on access network, core network equipment and long communica-tion paths between users and servers in the today’s cellu-lar networks. The cloud computing model can be used to avoid bottlenecks, better utilise available resources, and minimise delay. This paper investigates how the cloud computing model could be applied into the LTE (Long Term Evolution) cellular system and analyses how three deployed major cloud computing platforms could support this. This analysis showed that none of the studied cur-rently deployed cloud computing platforms can satisfy the derived requirements without enhancing and extending the functions and modules supported by each platform.
Keywords
Network virtualisation/cloudification, cellular systems, LTE
1.
INTRODUCTION
Cloud computing is one of the most discussed and promis-ing fields in modern computer science. Contrary to com-mon belief cloud computing is not a form of implementa-tion, but a model. Applying the cloud computing model to a system or part of a system can be denoted as cloudifica-tion. The implementation of the cloud computing model is typically denoted as virtualisation. The model used in this paper is defined by the National Institute for Standards and Technology (NIST) and states “Cloud computing is a model for enabling ubiquitous, convenient, on-demand network access to a shared pool of configurable comput-ing resources (e.g., networks, servers, storage, applica-tions, and services) that can be rapidly provisioned and re-leased with minimal management effort or service provider interaction.”, from [15]. The cloud computing model is mostly implemented in large data storage systems, provid-ing Infrastructure-as-a-Service (IaaS), Platform-as-a-Service (PaaS) and/or Software-as-a-Service (SaaS) service mod-els. A visual model of this can be seen in Figure 1, showing
Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. To copy oth-erwise, or republish, to post on servers or to redistribute to lists, requires prior specific permission and/or a fee.
17thTwente Student Conference on ITJune 25st, 2012, Enschede, The Netherlands.
Copyright2012, University of Twente, Faculty of Electrical Engineer-ing, Mathematics and Computer Science.
not only the essential characteristics and service models, but also the various deployment models.
Figure 1. Visual Model of Cloud Computing Def-inition, copied from [15]
IaaS provides customers processing, storage, network and other fundamental computing resources on which they can run software. This includes operating systems and appli-cations. This is done by providing the customer with a virtual machine. A virtual machine (VM) is a software implementation of a machine/computer and runs in much the same way as a physical machine. The idea is that mul-tiple VMs can run on a single physical machine and that the software running inside the virtual machine is limited to the resources and abstractions provided by the virtual machine and cannot break out of its virtualised environ-ment.
PaaS is using the services provided by IaaS, and offers customers an environment for application development, testing, deployment and hosting without the complexity of managing the underlying infrastructure and support-ing software. To facilitate the customer in such an envi-ronment the provider can support various programming languages, libraries, services and tools.
SaaS is using the services provided by PaaS and it de-livers its customers on-demand applications running on a cloud computing platform. The applications are accessi-ble, mostly over the Internet, through various clients like web browsers or thin clients. It is also possible to access the application through a programmable interface (API). SaaS customers do not have to be concerned about the un-derlying infrastructure and have no influence on the appli-cation features, apart from the user specific configuration settings.
Figure 1 shows that the cloud can be deployed in several ways. A private cloud is a cloud infrastructure operated solely for a single organization, whereas a public cloud is
publicly available and its services are free or offered on a pay-as-you-go model (services are only paid for the time that they are used). Community clouds can be placed somewhere in between, where the cloud infrastructure is shared between multiple organisations such that the costs are spread over more users than in the private cloud and less than a public cloud. A hybrid cloud is the combina-tion of two or more clouds, combining the advantages and disadvantages of those deployment models.
When providing IaaS, virtual machines are connected to each other in a virtual network. This is done by virtualis-ing network components like cables, switches, routers and firewalls. This virtual network can be made as simple or complex as one can imagine [17, 16]. Network virtualisa-tion in large date storage systems virtualises wired net-works. There is no practical need to virtualise a wireless network in such systems.
In the world of today, the increasing use of smartphones, tablets and other portable devices using the cellular sys-tem, requires from the mobile cellular system to cope with this increasing demand. A mobile cellular network can be considered as a radio network that is distributed over ra-dio coverage areas denoted as cells, which are usually con-trolled and served by fixed transmitter/receiver entities denoted as base stations. By connecting more cells to-gether, the cellular system can provide radio coverage over a wide geographic area. The part of the cellular system that encompasses the base stations is denoted as a radio access network (RAN) and the part that is encompassing the entities that are interconnecting the different base sta-tions and providing access to the Internet and other types of communication networks is denoted as Core Network. One of the most advanced cellular systems (i.e., 4th gen-eration) being currently under specification and standard-isation by 3GPP (Third Generation Partnership Project) is the Long Term Evolution (LTE) system, see Figure [1]. Today’s mobile cellular networks are highly centralised and not optimised for high-volume data applications, which will evolve with 4G (e.g., LTE) and beyond technologies. Operators typically use centralised network architectures that lead to very high bandwidth requirements on core net-work equipment and long communication paths between users and servers, which wastes network resources and in-creases delay. Shared distributed mobile network archi-tectures, are needed to avoid bottlenecks, better utilise available resources, and minimise delay.
The cloud computing model could be applied in mobile cel-lular systems to solve this issue by offering decentralised computing, smart storage, on-demand, elastic and Pay-as-you-Go services to third party operators and users, see e.g. [9]. There are a few papers available describing how virtual telecommunication operators are using the cloud comput-ing model to provide their services, see [4, 12]. There are also papers that focus on how wireless networking can be virtualised [27, 28], and even ones on the virtualisation of a cellular system, see e.g., [25, 26].
However, not many research activities focused on how ex-isting cloud computing platforms can be used to cloud-ify LTE cellular systems. This paper aims to resolve this dilemma. So the main research question is:
“How can the cloud computing model be applied in LTE cellular systems? ”
In order to answer the main research question the research is broken down in sub questions.
1. What is network cloudification?
2. Are network cloudification solutions already applied in LTE based cellular systems?
3. What are the main differences between applying net-work cloudification in (1) large data storage systems and (2) LTE cellular systems?
4. Which challenges have to be solved when using cur-rent cloud computing platforms in LTE based cellu-lar systems?
Research questions (1), (2) and (3) will be answered using literature study. The literature study will provide a list of challenges that current cloud computing platforms [8, 21, 22] need to resolve in order to apply the cloud computing model in LTE based cellular systems. Research question (4) will be answered by defining the main requirements needed to implement the cloud computing model in LTE based cellular systems and by studying how these require-ments are satisfied by the specifications of the currently deployed cloud computing platforms.
This paper is organized as follows. Section 2 gives an overview of network cloudification and its use in large data storage systems. Moreover this section answers research question (1). Section 3 gives an overview of the state art on applying network cloudification in cellular systems. This section also provides an answer to research question (2). Section 4 defines the main requirements needed to imple-ment the cloud computing model in LTE based cellular systems and analyses how these requirements are satis-fied by the specifications of the currently deployed cloud computing platforms. Furthermore, this section answers research question (4). Finally, section 5 concludes and provides recommendations for future work.
2.
NETWORK CLOUDIFICATION
The cloud computing model is used to maximise the utili-sation of large data storage systems, resulting in clustering and pooling of its resources and the use of virtual machines [3]. Whereas in the past, a single physical machine was used just as a single processing or storage machine, nowa-days it can host multiple virtual machines, each dedicated to its own task. This has the advantage of separating the physical infrastructure from its services and providing the service’s independence from its physical underlying hard-ware. This enables infrastructure providers to focus on the physical infrastructure and service providers to not be concerned about the physical limitations of the hardware. As a consequence, new business models and application areas will be enabled, see e.g., [15].
In large data storage systems, one or more virtual net-works interconnect the virtual machines to the physical outside world and to each other, enabling them to provide their services. These virtual networks consist of a set of virtualised networking resources like routers, switches and links, see e.g., [10].
To optimise and extend the utilisation of networking re-sources, the cloud computing model is applied in networks in the same way as it has been applied in large data stor-age systems. The process of applying the cloud computing concept in networks can be denoted as network cloudifica-tion, see e.g., [24].
Network cloudification can therefore be defined as the pro-cess of virtualising a network that can for example be com-posed of a radio access network, mobile core network, vir-tual local area network (VLAN), a virvir-tual private network
(VPN), active and programmable networks and overlay networks, see e.g., [17].
Figure 2. Network cloudification environment; based on [17]
Figure 2 shows the concept of network cloudification. On the bottom of the figure we can identify 3 users (U1, U2 and U3) each having their own infrastructure. The layout of the physical infrastructure can consist of e.g., comput-ers, communication links, data centres or any other en-tities imaginable. In this example, see Figure 2, user 1 is physically connected to the network operated by infras-tructure provider 1 (InP1) and users 2 and 3 are connected to the infrastructure provider 2 (InP2). Both infrastruc-ture providers support a physical network that consists of network entities and links connecting them. The infras-tructure providers are interconnected to each other using physical links in order to create a network of networks and exchange data between them. In this example, the infras-tructure providers are able to provide an IaaS virtualised services to the service providers, see Figure 2.
A service provider is able to provide a PaaS and/or SaaS environment that use the virtualised services provided by IaaS. The PaaS environment can be used for application development, testing, deployment and hosting without the complexity of managing the underlying infrastructure and supporting software, by using. The SaaS environment pro-vides its customers on-demand applications running on a cloud platform.
Multiple service providers can lease the networking infras-tructure resources (IaaS), as shown in Figure 2, to cre-ate and deploy multiple virtualised networks (PaaS and/or SaaS) on top of the physical infrastructure (IaaS). In this situation the service providers can create and de-ploy virtual networks on-demand, that are elastic (i.e. vir-tualised resources can be dynamically allocated and re-leased) and are based on a Pay-as-you-Go model (i.e., pay only for the resources that are allocated and used) to third party service operators and users [15, 17, 16, 18].
3.
NETWORK CLOUDIFICATION IN LTE
CELLULAR NETWORKS
Currently, several papers describe architectures that en-able the cloud computing concept to be applied in com-munication networks.
For example, [3] briefly discusses some results of the EU FP7 project SAIL [23], related to cloud networking that of-fers a combination and integration of cloud computing and virtual networking. In addition to SAIL [23] the following projects are considering network virtualisation concepts: GEYSERS [11] and OFELIA [20].
In [13] a concept is discussed that applies network shar-ing mechanisms usshar-ing virtualisation technologies to design a reconfigurable mobile network (RMN) architecture de-signed for flexibility and re-configurability. [5] discusses an IP Multimedia Subsystem (IMS) framework with cloud computing architecture for use in high quality multime-dia applications. In [4] a Virtual Telco is discussed that uses a Cloud computing approach in order to replace the costly dedicated hardware. This is realised by implement-ing several centralised control plane functions and other services with distributed solutions that may be allocated on-demand over a pool of dependable, dynamically con-tracted computing and networking resources that are easy to manage. In [14] an algorithm is discussed that is used for the placement of virtualised service components to net-work sites, where the performance metric is the cost for acquiring components between the sites. The resulting optimisation problem, is referred to as the k-Component Multi-site Placement Problem, and is applicable to service distribution in a wide range of communication networking scenarios.
[27] discusses the distributed wireless communication sys-tem (DWCS), which is a virtualised architecture for a wire-less access system with distributed antennas, distributed processors, and distributed control. When distributed an-tennas are used, the system capacity can be expanded through dense frequency reuse, and the transmission power can be greatly decreased. Moreover, by using distributed processors control, the system works like a software or network radio. In this way, different standards can exist, and the system capacity can be increased by co-processing of signals to and from multiple antennas. [28] introduces the a TDD (Time Division Duplex) WiMAX SDR (Software Defined Radio) Base Station implemented on a commodity server, in conjunction with a novel de-sign of a remote radio head (RRH). [28] also presents a working prototype of a virtual BS (VBS) pool, exploring the systems challenges in supporting a VBS pool on multi-core IT platforms. Please note that [28] emphasises that the virtual base stations are implemented using the Open Base Station Architecture Initiative (OBSAI) [19] and the Common Public Radio Interface (CPRI) [7, 6].
The research activities mentioned above are describing possible ways of how network cloudification can be used in wireless networks. However, it is not clear whether these architectures are able to provide the same functional and performance capabilities as the ones provided by non-virtualised LTE networks.
Figure 3. LTE Evolved Packet System (EPS) Ar-chitecture, based on [2]
Figure 3 shows a very basic architecture of the LTE Evolved Packet System (EPS), which is designed to handle the data traffic efficiently (from a performance and costs perspec-tive) by avoiding protocol conversion and minimising the amount of network nodes the data needs to pass.
The User Equipment (UE), e.g., smartphones and other 4G enabled devices, is connected to the LTE core network over the LTE radio access network, which consists of base transceiver stations, denoted as eNodeB (evolved NodeB). An eNodeB is responsible for accessing the radio channel and scheduling the air interface resources between the UE and establishing a point-to-point link between the UE and the LTE core network. In LTE the radio access network is denoted as e-RAN (evolved - RAN) and the core network is denoted as EPC (Evolved Packet Core).
The EPC is composed of five network elements: the S-GW (Serving GW), the PDN GW (Packet Data Network Gate-way), the MME (Mobility Management Entity), PCRF (Policy and Charging Control Element) and the HSS (Home Subscriber System).
Home Subscriber System (HSS): Is a database that contains user-related and subscriber-related informa-tion. It also provides support functions for mobility management, call and session setup, user authenti-cation and access authorisation.
Serving Gateway (S-GW): It supports the transport of the user data between the UE and the external networks. The Serving GW connects the e-RAN and the EPC, routes the incoming and outgoing data and ensures a smooth handover between eNodeBs. The serving GW is connected to the PDN GW.
Packet Data Network Gateway (PDN GW): The PDN GW connects the EPC to other external net-works. The PDN GW routes packets to and from the PDNs (Packet Data Networks). The PDN GW also performs various functions such as IP address / IP prefix allocation or policy control and charging. The two gateways (S-GW and PDN-GW) are spec-ified independently but they can also be collocated. Both gateways (S-GW and PDN GW) support the user plane (i.e., user data).
Mobility Management Entity (MME): is supporting only the control plane (i.e., signalling and control data). It handles the signalling related to mobility and security for the e-RAN. The MME is responsi-ble for the tracking and the paging of UE when is operating in idle-mode.
Policy and Charging Control Element (PCRF): it supports the control plane, and it provides features like policy control and charging control.
The EPC is connected to (1) the Operator’s services, which can include the IP Multimedia Subsystem (IMS) and can usually run on data centres and (2) external networks (like the Internet). The IMS is an LTE service platform to pro-vide IP multimedia services to the LTE subscribers. The EPS architecture separates the user data (user plane) and the signalling (control plane). Thanks to this func-tional split, the operators can dimension and adapt their network easily. The LTE network is an all-IP system, which means that data is carried in IP packets avoiding the need for conversion and so enabling the fast connectivity with other networks like the Internet.
To apply the cloud computing model in a LTE network infrastructure, virtualisation of all above mentioned com-ponents (e-RAN, EPC and Operators services platforms, like IMS (i.e., data centres)) is needed, enabling multiple mobile network operators to create their own virtual net-work on the same infrastructure [12, 25, 28]. Since over
80% of the infrastructure is software based, most of the in-frastructure is already cloud computing enabled and can be virtualised by using e.g. virtual machines. So virtu-alising, for example, the eNodeB in the e-RAN could be similar to server virtualisation as is done in large data storage systems [26].
In [25], network virtualisation is applied in Long Term Evolution (LTE) and specifies a virtualised LTE frame-work that realises an auctioned based dynamic spectrum sharing.
Figure 4. LTE eNodeB virtualisation framework architecture; copied from [25]
The physical resources need to be shared between multiple virtual instances, which is done by a so called hypervisor and sits on top of the physical hardware to schedule the re-sources. This results in the fact that the physical eNodeB is virtualised into a number of virtual eNodeBs which can be seen in Figure 4. In addition, the hypervisor is respon-sible for scheduling the air interface resources between the virtual eNodeBs running on top. The “Spectrum configu-ration and Bandwidth estimation” in Figure 4 is respon-sible for allocating the spectrum that the virtual eNodeB is supposed to operate in, and as well as estimating the required bandwidth of the operator. The “Spectrum al-location unit” is responsible for scheduling the spectrum among the different virtual eNodeBs [25].
Figure 4 shows the virtualisation of the eNodeB and e-RAN to be used by multiple operators. Virtualising the EPC and the Operators service platform, e.g., IMS, can be done in the same manner and requires the virtualisation of the EPC and IMS entities, respectively. These entities will then be virtualised and can run on a common (or separated) cloud computing centres.
By virtualising the e-RAN, EPC and the Operators service platform, a cloudified LTE can be developed that is able to automate, manage and orchestrate the pool of resources to deliver cloud computing services. The cloudification of the LTE system requires one or more cloud computing platforms to cloudify the e-RAN, EPC and the Operator’s services infrastructures. In this way the LTE virtual net-works that are elastic and are based on a Pay-as-you-Go model can be enabled on-demand, to third party LTE ser-vice operators and users.
4.
APPLYING EXISTING CLOUD
COM-PUTING PLATFORMS IN LTE SYSTEMS
4.1
Research Methodology
In order to analyse cloud computing platforms for their ability to be used in cloudifying LTE systems, criteria need
to be established, see [9]. The criteria are setup to measure the cloud computing platforms ability to cloudify the LTE system as described section 3. By researching the cloud computing platforms documentation, it will become clear whether these criteria can be met or not. These criteria are:
1. Support virtualised foundational infrastructural re-sources from the Radio Access Network and Mobile Core Network and Data Centre
2. Support a coordination and orchestration function from the infrastructure perspective for the composi-tion, provisioning, life cycle management of services and maintenance of agreed SLA guarantees spread across Radio Access Network and Core Network. 3. Support monitoring facilities for extraction,
prepro-cessing, distribution, storage, analysis and notifica-tion of metrics such that higher-level services can utilise these facilities to enable adaptive platform ser-vices (such as SLAs), where certain guarantees and service quality (QoS) and/or experience (QoE) must be maintained
4. Support mechanisms to deliver virtual resources as service for the execution of non-traditional virtu-alised instances such as virtuvirtu-alised eNodeB, EPC (e.g., MME, PGW, SGW) and IMS functionality 5. Support unified and consistent interfaces on top of
and between disparate cloud computing physical in-frastructure management frameworks (i.e., Radio Ac-cess Network, Mobile Core Network, Data centre) The analysis will be done using the documentation of the latest available version of the cloud computing platforms, see Table 1.
Table 1. Cloud computing platform version and release date
Platform Version Release date OpenStack [22] 2012.1 (Essex) 04-05-2012 Eucalyptus [8] 2.0.3 25-05-2011 OpenNebula [21] 3.4.1 (Wild Duck) 03-05-2012
4.2
Cloud computing platforms
4.2.1
OpenStack
OpenStack Essex, [22] the fifth version of OpenStack, focuses on quality, usability and extensibility across enterprise, service provider and high performance comput-ing deployments. OpenStack Es-sex allows users to leverage pools of on-demand, self-managed com-pute, storage and networking
re-sources to build efficient, automated private and public cloud infrastructures.
OpenStack Essex includes, but is not limited to, the fol-lowing modules:
Compute: The OpenStack cloud operating system en-ables service providers to offer on-demand computing resources, by provisioning and managing large net-works of virtual machines. Compute resources are
accessible via APIs for developers building cloud ap-plications and via web interfaces for administrators and users. The Compute architecture is designed to scale horizontally. OpenStack is designed to pro-vide flexibility and to manage and automate pools of Compute resources and can work with multiple supported hypervisors in a virtualised environment like KVM (Kernel-based Virtual Machine) and Xen. Linux container technology such as LXC (Linux Con-tainers) is also supported for scenarios where users need to minimise virtualisation overhead and gain greater efficiency and performance. In addition to different hypervisors, OpenStack supports ARM (Ad-vanced RISC-Reduced Instruction Set Computer Ma-chine) and alternative hardware architectures. Object Storage: OpenStack provides redundant,
scal-able object storage using clusters of standardized servers capable of storing petabytes of data. Object Storage is not a traditional file system, but rather a distributed storage system for static data such as virtual machine images, photo storage, email stor-age, backups and archives. Having no central ”brain” or master point of control provides greater scalabil-ity, redundancy and durability. Objects and files are written to multiple disk drives spread across servers in the data centre, with the OpenStack software re-sponsible for ensuring data replication and integrity across the cluster. Storage clusters scale horizontally simply by adding new servers.
Block Storage: OpenStack also provides persistent block level storage devices for use with OpenStack com-pute instances. The Block Storage system manages the creation, attaching and detaching of the block de-vices to servers. Block Storage volumes are fully inte-grated into OpenStack Compute and the Dashboard allowing for cloud users to manage their own storage needs. In addition to using simple Linux server age, it has unified storage support for numerous stor-age platforms. Block Storstor-age is appropriate for per-formance sensitive scenarios such as database stor-age, expandable file systems, or providing a server with access to raw block level storage.
Networking: OpenStack Networking is a pluggable, scal-able and API-driven system for managing networks and IP addresses. OpenStack provides flexible net-working models to suit the needs of different appli-cations or user groups. Standard models include flat networks or VLANs for separation of servers and traffic. OpenStack Networking manages IP ad-dresses, allowing for dedicated static IPs or DHCP (Dynamic Host Configuration Protocol). Floating IPs allow traffic to be dynamically rerouted to any of the Compute resources, which allows to redirect traf-fic during maintenance or in the case of failure. Users can create their own networks, control traffic and connect servers and devices to one or more networks. Administrators can take advantage of software-defined networking (SDN) technology like OpenFlow to al-low for high levels of multi-tenancy and massive scale. OpenStack Networking has an extension framework allowing additional network services, such as intru-sion detection systems (IDS), load balancing, fire-walls and virtual private networks (VPN) to be de-ployed and managed.
Dashboard: The OpenStack Dashboard provides admin-istrators and users a graphical interface to access,
provision and automate cloud-based resources. The extensible design makes it easy to plug in and expose third party products and services, such as billing, monitoring and additional management tools. De-velopers can automate access or build tools to man-age their resources using the OpenStack API or the Amazon EC2 (Elastic Compute Cloud) compatibil-ity API. The dashboard is an extensible web app that allows cloud administrators and users to control their compute, storage and networking resources. As a cloud administrator, the dashboard provides an over-all view of the size and state of the cloud and is able to create users and projects, assign users to projects and set limits on the resources for those projects. The Dashbord provides users a self-service portal to provision their own resources within the limits set by administrators.
4.2.2
Eucalyptus
Eucalyptus [8] is a widely de-ployed software platform for
pri-vate IaaS clouds. It uses existing infrastructures to create a scalable, secure web services layer that abstracts com-pute, network and storage to offer IaaS. Eucalyptus takes advantage of modern infrastructure virtualisation software to create elastic pools that can be dynamically scaled up or down depending on application workloads. Eucalyptus web services are uniquely designed for hybrid clouds using the industry standard Amazon Web Services (AWS) API. Eucalyptus key functionality includes:
Self-service User Portal: The Eucalyptus User Console provides an interface for users to self-service provi-sion and configure compute, network, and storage resources. Development and test teams can flexi-bly and securely manage their virtual instances us-ing built-in key management and encryption capa-bilities. Access to virtual instances is available us-ing familiar SSH (Secure Shell) and RDP (Remote Desktop Protocol) mechanisms. Virtual instances with application configuration can be stopped and re-started securely using encrypted boot from Elas-tic Block Storage (EBS) capability.
High Availability Cloud IaaS: Systems failure is ex-pected and unavoidable in an data center, but Eu-calyptus has been designed to be highly available at multiple levels. IaaS service components Cloud Controller (CLC), Cluster Controller (CC), Walrus, Storage Controller (SC) and VMware Broker (VB) are configurable as redundant systems that are very resilient to multiple types of failures. Management state of the cloud machine will be preserved and re-verted to normal operating conditions in the event of a hardware or software failure.
Heterogeneous Hypervisor Management: Build and manage multiple hypervisor cluster environments in an IaaS cloud. Manage existing VMware vSphere, VMware ESX (Elastic Sky X), VMware ESXi, KVM, and Xen virtual environments from a single pane as Amazon AWS compatible Eucalyptus hybrid clouds. Manage Multiple Machine Image Formats:
Euca-lyptus can run multiple versions of Microsoft Win-dows and Linux virtual machine images on IaaS clouds and build a library of Eucalyptus Machine Images (EMIs) with application metadata that are decou-pled from infrastructure details to allow them to run
on Eucalyptus clouds. Amazon Machine Images are also compatible with Eucalyptus clouds. VMware Images and VMware vApps can be converted to run on Eucalyptus clouds, and Amazon AWS and Ama-zon compatible public clouds.
Scriptable Architecture: Production environments want flexible, configurable and scriptable system to auto-mate as many functions as possible in a cloud man-agement platform. Users have the choice of inte-grating web services API with their preferred ven-dor management systems or using an Amazon AWS compatible management stack to manage Eucalyp-tus clouds. All the functionality is exposed in a scriptable interface to work with other system man-agement tools.
Identity Management: Eucalyptus user identity man-agement can be integrated with existing Microsoft Active Directory or LDAP (Lightweight Directory Access Protocol) systems to have fine-grained role based access control over cloud resources. Flexible Accounting, Chargeback, and Quota Management Define and allocate resource quotas for the users and groups with Eucalyptus 3’s quota management fea-tures. Control resource allocation across clusters, defined by users and groups.
Accounting, Chargeback, and Quota Management: Define and allocate resource quotas for users and groups with quota management features. Control resource allocation across clusters, defined by users and groups and analyse cloud usage patterns by us-ing resource accountus-ing module. Compute and stor-age usstor-age data are available in various formats for vi-sualization, reporting, and analysis by business sys-tems for both enterprises and integration with charge-back and billing platforms.
Resource Administration And Console: The Eucalyp-tus Dashboard provides cloud administrators with a graphical console for performing several cloud man-agement tasks including all virtual and physical re-source management and virtual cloud rere-source con-figuration, provisioning and de-provisioning.
4.2.3
OpenNebula
OpenNebula [21] is a cloud com-puting toolkit for managing
het-erogeneous distributed data centre infrastructures. The OpenNebula toolkit manages a data centre virtual infras-tructure to build private, public and hybrid IaaS clouds. OpenNebula orchestrates storage, network, virtualisation, monitoring, and security technologies to deploy multi-tier services (e.g. compute clusters) as virtual machines on distributed infrastructures, combining both data centre resources and remote cloud resources, according to allo-cation policies.
OpenNebula key features include:
User Security Management: Secure and efficient users and groups pluggable subsystem for authentication and authorisation of requests with complete func-tionality for user management. The authorisation framework allows multiple-role support for different types of users and administrators, delegated control to authorised users, secure isolated multi-tenant en-vironments, and easy resource (VM template, VM image, VM instance, virtual network and host) shar-ing. Administrators have complete functionality for
management of grouped users where each group has configurable access to shared resources so enabling a multi-tenant environment with multiple groups shar-ing the same infrastructure.
On-demand Provision of Virtual Data Centres: A Virtual Data Centres (VDC) is a fully-isolated virtual infrastructure environment where a group of users, under the control of the VDC administrator, can create and manage compute, storage and net-working capacity. There is also support for the cre-ation and management of multiples VDCs within the same logical cluster and zone.
Control and Monitoring of Virtual Infrastructure: OpenNebula has a Storage and Repository Subsys-tem with functionality for VM image and Subsys-template management. The platform also has full control of VM instance life-cycle and functionality for VM in-stance management. OpenNebula also has broad network virtualisation capabilities with traffic isola-tion, ranged or fixed networks, definition of generic attributes to define multi-tier services consisting of groups of inter-connected VMs, and functionality for virtual network management to interconnect VM in-stances. The tagging of users, VM images and vir-tual networks with arbitrary metadata that can be later used by other components and API’s future use. Virtual Machine Configuration: Complete definition of VM attributes and requirements and the support for automatic configuration of VMs with advanced contextualization mechanisms with a wide range of guest operating system including Microsoft Windows and Linux. For networking, OpenNebula offers flexi-ble network definition including configuration of fire-walls for VMs to specify a set of black/white TCP / UDP ports.
Control and Monitoring of Physical Infrastructure: OpenNebula is able to deploy public, private and hy-brid clouds. The Host Management Subsystem of-fers functionality for management of physical hosts with the dynamic creation of clusters as a logical set of physical resources, namely: hosts, networks and data stores, within each zone. OpenNebula has a powerful and extensible built-in monitoring sub-system and resource quota management to allocate, track and limit resource utilisation. The virtualisa-tion Subsystem supports a wide range of hypervi-sors like Xen, KVM and VMware ESX, centralised management of environments with multiple hypervi-sors, and support for multiple hypervisors within the same physical box. Storage Subsystem with support for multiple data stores to balance I/O operations between storage servers, or to define different SLA policies (e.g. backup) and performance features for different VM types or users.
Standard Cloud Interfaces and Self-Service Portal: OpenNebula is able to simultaneously expose mul-tiple cloud APIs including the AWS EC2 API ser-vice for compatibility with EC2 ecosystem tools and client tools. OpenNebula also has a self-service pro-visioning portal to allow non-IT end users to eas-ily create, deploy and manage compute, storage and network resources and a Unix-like Command Line Interface to manage all resources: users, VM im-ages, VM templates, VM instances, virtual networks,
zones, VDCs, physical hosts, accounting, authenti-cation, authorisation for cloud administrators. The modular and extensible Graphical Interface provid-ing usage graphics and statistics with cloudwatch-like functionality, VNC support, different system views for different roles, catalogue access, multiple-zone management.
4.3
Analysis and challenges
The analysis starts by mapping the features and function-ality of the cloud computing platforms to the criteria intro-duced in Section 4.1. The analysis can be found in Table 2, 3 and 4 showing the criteria and the modules which could be used to support each criterion. The criteria are numbered according to the numbering used in Section 4.1. It is important to be noticed that the functions/modules (including the APIs) supported by the three cloud com-puting platforms are able to support wired and fixed vir-tualisation scenarios that are making use of large data centres.
These functions/modules (including the APIs) are not de-signed and implemented to support virtualised scenarios where the allocation and release of e-RAN, EPC and Op-erators’ services (e.g., IMS) resources takes place in a very dynamic and fast changing environment.
Therefore, in order to support the five criteria described in Section 4.1, see Table 5, these functions/modules (in-cluding the APIs) need to be enhanced such that they can support challenges, such as dynamic network access connectivity and release, QoS support, cross-operator dy-namic mobility and roaming management, security and privacy support, billing and charging using a Pay-as-you-go policy for an environment where the allocation and re-lease of e-RAN, EPC and Operators’ services (e.g., IMS) resources takes place in a very dynamic and fast changing environment.
It needs to be noted that depending on the operators re-quirements, the use of additional not in this paper dis-cussed functions/modules could be used. However, those additional modules probably also need to be enhanced in order to support the criteria given in Section 4.1. Table 2, 3 and 4 show the criteria and the modules of OpenStack, Eucalyptus and OpenNebula, respectively, which need to be enhanced in order to support these criteria.
Table 2. Analysis OpenStack Criterion Modules to be enhanced
1 Compute and Networking module 2 Compute and Networking module 3 Dashboard
4 Compute and Networking module 5 Dashboard
Note that all three cloud computing platforms have sup-port for billing and charging using a Pay-as-you-go policy which can be integrated by their API into other financial systems of the operator.
The analysis is summarised in Table 5 and clearly shows how the cloud computing platforms compare to each other and whether or not criteria have fully be met or not. Table 2, 3, 4 and 5 show that enhancements and exten-sions need to be made in order to support challenges, such as dynamic network access connectivity and release, QoS support, cross-operator dynamic mobility and roam-ing management, security and privacy support, billroam-ing and
Table 3. Analysis Eucalyptus Criterion Could be supported by
1 Heterogeneous Hypervisor Management, Manage Multiple Machine Image Formats and High Availability Cloud IaaS
2 Self-service User Portal, Resource Admin-istration And Console and Identity Man-agement
3 Self-service User Portal, Resource Admin-istration And Console and Identity Man-agement
4 Scriptable Architecture, Heterogeneous Hy-pervisor Management, Manage Multiple Machine Image Formats and High Avail-ability Cloud IaaS
5 Scriptable Architecture, Self-service User Portal, Resource Administration And Con-sole and Identity Management
Table 4. Analysis OpenNebula Criterion Could be supported by
1 Control and Monitoring of Physical Infras-tructure and On-demand Provision of Vir-tual Data Centres
2 Control and Monitoring of Virtual Infras-tructure and Virtual Machine Configura-tion
3 User Security Management and Standard Cloud Interfaces and Self-Service Portal 4 On-demand Provision of Virtual Data
Centres, Control and Monitoring of Virtual Infrastructure and Virtual Machine Con-figuration
5 User Security Management and Standard Cloud Interfaces and Self-Service Portal
charging using a Pay-as-you-go policy for an environment where the allocation and release of e-RAN, EPC and Op-erators’ services (e.g., IMS) resources take place in a very dynamic and fast changing environment for all three cloud computing platforms.
5.
CONCLUSIONS AND FUTURE WORK
This paper investigated how the cloud computing model could be applied into the LTE (Long Term Evolution) cel-lular system and analyses how three deployed major cloud computing platforms could support this.
Section 2 answered research question (1) by providing an overview of network cloudification and its use in large data storage systems. Section 3 answered research question (2) by providing an overview of the state art on applying network cloudification in cellular systems. Section 4 an-swers research questions (3) and (4) by defining the main requirements needed to implement the cloud computing model in LTE based cellular systems and analyses how these requirements are satisfied by the specifications of the currently deployed cloud computing platforms. In particular, this paper showed that none of the three major current cloud computing platforms fully satisfy the criteria from Section 4.1, as the most of the existing func-tions/modules cannot support the challenges that need to be fulfilled by scenarios where the allocation and release of e-RAN, EPC and Operators’ services (e.g., IMS)
re-Table 5. Analysis results
```` ```` ``` Criteria Platform OS1 Eu2 ON3 1 4 5 6 2 4 5 6 3 4 5 6 4 4 5 6 5 4 5 6 1OpenStack [22] 2 Eucalyptus [8] 3 OpenNebula [21]
4Cannot be satisfied, at least modules shown in Table 2 need to be enhanced and extended 5
Cannot be satisfied, at least modules shown in Table 3 need to be enhanced and extended 6
Cannot be satisfied, at least modules shown in Table 4 need to be enhanced and extended
sources takes place in a very dynamic and fast changing environment.
It is expected that these criteria could be satisfied if, at least, the functions/modules listed in Table 2, 3, 4 are en-hanced in order to fulfil challenges, such as dynamic net-work access connectivity and release, QoS support, cross-operator dynamic mobility and roaming management, se-curity and privacy support, billing and charging using a Pay-as-you-go policy for an environment where the alloca-tion and release of e-RAN, EPC and Operators’ services (e.g., IMS) resources takes place in a very dynamic and fast changing environment.
The functions/modules that must be enhanced are as fol-lows: • OpenStack – Compute – Networking – Dashboard • Eucalyptus
– Heterogeneous Hypervisor Management – Manage Multiple Machine Image Formats – High Availability Cloud IaaS
– Self-service User Portal
– Resource Administration And Console – Identity Management
– Scriptable Architecture • OpenNebula
– Control and Monitoring of Physical Infrastruc-ture
– On-demand Provision of Virtual Data Centres – Control and Monitoring of Virtual
Infrastruc-ture
– Virtual Machine Configuration – User Security Management
– Standard Cloud Interfaces and Self-Service Por-tal
The functions/modules of the cloud computing platforms OpenStack, Eucalyptus and OpenNebula could satisfy sev-eral of the derived criteria after solving sevsev-eral challenges. Once enhancements have been made in the above specified functions/modules, the enhanced cloud computing plat-forms could be used to enable realisation of the cloudifi-cation of the LTE cellular system.
6.
REFERENCES
[1] 3GPP. LTE specifications.
http://www.3gpp.org/lte, visited in June 2012. [2] 3GPP TS 23.401. General Packet Radio Service
(GPRS) enhancements for Evolved Universal Terrestrial Radio Access Network (E-UTRAN) access (Release 11). Technical Specification TS 23.401, V11.1.0, 3GPP, March 2012.
[3] B Ahlgren, P A Aranda, P Chemouil, S Oueslati, L M Correia, H Karl, M Sllner, and A Welin. Content, connectivity, and cloud: Ingredients for the network of the future. IEEE Communications Magazine, 49(7):62–70, 2011.
[4] P Bosch, A Duminuco, F Pianese, and T L Wood. Telco clouds and Virtual Telco: Consolidation, convergence, and beyond. In Proceedings of the 12th IFIP/IEEE International Symposium on Integrated Network Management, IM, pages 982–988, 2011. [5] J.L. Chen, S.L. Wuy, Y.T. Larosa, P.J. Yang, and
Y.F. Li. IMS cloud computing architecture for high-quality multimedia applications. In Wireless Communications and Mobile Computing Conference (IWCMC), 2011 7th International, pages 1463–1468. IEEE, 2011.
[6] Common Public Radio Interface (CPRI). CPRI Specification. Interface Specification V4.0, 2006. [7] Common Public Radio Interface (CPRI). Official
website. http://www.cpri.info, visited in June 2012.
[8] Eucalyptus. Cloud Computing Software.
http://www.eucalyptus.com, visited in June 2012. [9] European FP7 Integrated Project. Mobile Cloud
Networking. FP7-ICT-2011-8, starts in November 2012.
[10] A Galis, S Clayman, L Lefevre, A Fischer, H De Meer, J Rubio-Loyola, J Serrat, and S Davy. Towards in-network clouds in future internet. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), 6656:19–33, 2011. [11] Generalised Architecture for Dynamic Infrastructure
Services, (GEYSERS). EU FP7 official website. http://www.geysers.eu, visited in June 2012. [12] S Hall. Evolution of telco services utilising
Infrastructure-as-a-Service (IaaS). In 2011 15th International conference on Intelligence in Next Generation Networks, ICIN 2011, pages 247–252, 2011.
[13] A Khan, D Jurca, K Kozu, W Kellerer, and M Yabusaki. The Reconfigurable Mobile Network. Proc. of IEEE International conference on
Communications Workshops, ICC 2011, pages 1–5, 2011.
[14] J. Leblet, Z. Li, G. Simon, and D. Yuan. Optimal network locality in distributed virtualized data-centers. Computer Communications, 34(16):1968–1979, 2011.
[15] Peter Mell and Timothy Grance. The NIST Definition of Cloud Computing. Special Publication
800-14, National Institute of Standards and Technology (NIST), September 2011.
[16] N. M. Mosharaf Kabir Chowdhury and Raouf Boutab. A survey of network virtualization. 2009. [17] N. M. Mosharaf Kabir Chowdhury and Raouf
Boutab. Network Virtualization: State of the Art and Research Challenges. IEEE Communications Magazine, pages 20–26, July 2009.
[18] Norbert Niebert, Ibtissam El Khayat, Stephan Baucke, Ralf Keller, Ren´e Rembarz, and Joachim Sachs. Network Virtualization: A Viable Path Towards the Future Internet. Wireless Pers Commun, 45:511–520, March 2008.
[19] Open Base Station Architecture Initiative (OBSAI). Official website. http://www.obsai.org, visited in June 2012.
[20] OpenFlow in Europe, (OFELIA). EU FP7 Official website. http://www.fp7-ofelia.eu, visited in June 2012.
[21] OpenNebula. The Open Source Solution for Data Center Virtualization. http://www.opennebula.org, visited in June 2012.
[22] Rackspace Cloud Computing. OpenStack Open Source Cloud Computing Software.
http://www.openstack.org, visited in June 2012. [23] Scalable and Adaptive Internet Solutions (SAIL).
EU FP7 Pfficial website.
http://www.sail-project.eu, visited in June 2012. [24] P Schoo, V Fusenig, V Souza, M Melo, P Murray,
H Debar, H Medhioub, and D Zeghlache. Challenges for cloud networking security. HP Laboratories Technical Report, pages 1–17, 2010.
[25] Y Zaki, M A Khan, L Zhao, and C G¨org. Realizing the Broker Based Dynamic Spectrum Allocation through LTE Virtualization and Uniform Auctioning. Proc. of IFIP NETWORKING 2011 Workshop, 6827 LNCS:85–97, 2011.
[26] L Zhao, M Li, Y Zaki, A Timm-Giel, and C G¨org. LTE virtualization: From theoretical gain to practical solution. In Proceedings of the 23rd International Teletraffic Congress, ITC 2011, pages 71–78, 2011.
[27] S Zhou, M Zhao, X Xu, J Wang, and Y Yao. Distributed wireless communication system: A new architecture for future public wireless access. IEEE Communications Magazine, 41(3):108–113, 2003. [28] Z Zhu, P Gupta, Q Wang, S Kalyanaraman, Y Lin,
and H Franke. Virtual base station pool: Towards a wireless network cloud for radio access networks. In Proceedings of the 8th ACM International conference on Computing Frontiers, CF’11, 2011.