Overview of Virtualization for Small Cells

SMALL CELL FORUM

RELEASE

7. 0

scf.io

17 : : 2 25 5

E S I R P R E T

H O OM E M N E E

U R RB A B N

R U R A L & R E M O T E V I I R RT U T A U L I I Z ZA T I I O ON

DOCUMENT

106.07.01 Vir tualization for small cells: Overview Over view

June 2015

Solving the HetNet puzzle www.scf.io/

www.smallcellforum.org

SMALL CELL FORUM

RELEASE

7. 0

Small Cell Forum accelerates small cell adoption to drive the wide-scale adoption of small cells and accelerate the delivery of integrated HetNets. We are not a standards organization but partner with organizations that inform and determine standards development. We are a carrier-led organization. This means our operator members establish requirements that drive the activities and outputs of our technical groups. We have driven the standardization of key elements of small cell technology including Iuh, FAPI/SCAPI, SON, the small cell services API, TR-069 evolution and the enhancement of the X2 interface. Today our members are driving solutions that include small cell/Wi-Fi integration, SON evolution, virtualization of the small cell layer, layer, driving mass adoption via multi-operator neutral host, ensuring a common approach to service APIs to drive commercialisation and the integration of small cells into 5G standards evolution. The Small Cell Forum Release Program has now established business cases and market drivers for all the main use cases. This document is part of Release 7: HetNet and SON. Small Cell Forum denes HetNet as a ‘multi-x environment – multi-technology, multi-technology, multi-domain, multi-spectrum, multi-operator and multi-vendor. multi-vendor. It must be able to automate the reconguration of its operation operation to deliver assured service quality across the entire network, and exible enough enough to accommodate accommodate changing user needs, business goals and subscriber behaviors.’ behaviors.’ Small Cell Forum Release website can be found here: her e: www.scf.io www.scf.io All content in this document including links and references are for informational informational purposes only and is provided “as is” is” with no warranties whatsoever including any warranty of merchantability, tness for any particular purpose, or any warranty otherwise arising out of any proposal, specication, or sample. No license, express or implied, to any intellectual property rights is granted or intended hereby. hereby.

If you would like more information about Small Cell Forum or would like to be included on our mailing list, please contact: Email [email protected] Post Small

Cell Forum, PO Box 23, GL11 5WA UK

Member Services

[email protected] [email protected] g

scf.io

Executive summary In June 2014, the SCF’s Operator Group tasked the Forum with performing a comprehensive analysis into the role of small cell virtualization. This overview document describes the findings of that activity. Specifically, it is recognized that there are key benefits and drivers towards centralization and virtualization of the small cell RAN. These include: •

•

•

•

•

improved coordination of the radio that then delivers higher performance and efficiencies; enhanced scalability of small cell deployments with simplified management of the many physical small cells, reduced cost from being able to lower peak-to-mean ratios for compute resources, accelerated upgrade lifecycles enabling new features to be deployed on a centralized virtual platform; and flexibility that enables optimum work load placement according to the availability of compute and transport resources.

However, there is a cost associated with the benefits gained from centralization of network functions, specifically related to the upgraded transport requirements. Examining the two extremes indicates that the well know BBU/RRH split based on CPRI/ORI may require costly dark fiber operated over short distances, compared with the conventional distributed RAN based S1 solutions that can be backhauled over nationwide low cost IP networks. Hence, i n general, splitting the functionality of the eNB at lower layers, generally increases the benefits available (in particular, from a radio perspective), but also requires higher performance transport, which has higher cost. The SCF have analyzed several different ‘functional split’ points in-between these two extremes, to identify a sweet spot, which has most of the benefits without significantly increasing the transport performance requirements (and hence costs). The main focus of the Small Cell Forum’s virtualization study has therefore been to look more broadly at the virtualization topic compared with previous CPRI/ORI approaches, to be able to quantify the costs and benefits of different ‘functional splits’ and find the optimum way to realize small cell virtualization: •

•

•

•

•

[SCF159] Functional Splits and Use Cases [ 1]: describes the different functional split options and their respective transport requirements [SCF160] Capacity and Coverage [ 2]: describes the centralization benefits of the different splits from an RF perspective [SCF158] Business Case [3]: Evolves the SCF’s Urban Business Case baseline to quantify the TCO increase based on transport costs of different functional splits [SCF161] Network Aspects [4]: Describes the wider system impacts of small cell virtualization [SCF106] [5] Pulls it all together into a concise overview

Specifically, the analysis concludes that the MAC/PHY split delivers most of the benefits of centralization, with only a small increase in transport performance and is well aligned with the current small cell multi-vendor ecosystem approach based on the Functional Application Platform Interface (FAPI). For those use cases that face restrictions from a transport perspective such that th e backhaul system cannot be enhanced to support the MAC/PHY split, then the PDCP/RLC split has the potential to

Report title: Virtualization for small cells: Overview Issue date: 09 June 2015 Version: 106.07.01

deliver a subset of the centralization benefits b ut with no additional transport requirements when compared with conventional small cell deployments. Given such a positive analysis, the Forum is motivated to define a transportable interface for the spit small cell, to establish a scalable ecosystem with a converged approach to virtualization across multiple suppliers.


Contents 1.

Introduction .....................................................................1

1.1 1.2

Structure of this document ................................................... 3 Value of SC-virtualization content for different audiences ......... 3

1.3 2. 3. 4.

Terminology ....................................................................... 3 Motivation for small cell virtualization study ....................5 Small cell decomposition: virtualization use cases ...........9 Coverage, capacity and performance aspects of small cell virtualization ............................................................ 12 Impact of base station virtualization on the end-to-end small cell system ............................................................14 Examining the business case impacts for small cell virtualization ..................................................................17 Virtualization study conclusions .....................................19

5. 6. 7.

8. Summary ........................................................................21 References ................................................................................23 Tables Table 3-1

Latency and bandwidth characteristics associated with these decomposition use cases. ............................................................... 10

Table 3-2

Fronthaul transport options ............................................................. 11

Table 4-1

Mapping of enhanced co-ordination techniques to virtualization use cases splits ................................................................................... 13


Figures Figure 1-1

Contrasting evolutions of the RAN architecture ................................... 1

Figure 2-1

Analysis of market adoption of RAN virtualization and small cells .......... 5

Figure 2-2

Factors identified by MNOs as the key barriers to adopting virtualization in the access network ................................................... 6

Figure 2-3

Complex issue required to be addressed when studying small cell virtualization .................................................................................. 8

Figure 3-1

Various possible LTE base station decompositions ............................... 9

Figure 3-2

Hosting options for large enterprise deployment of virtualized small cell function ..................................................................................10

Figure 3-3

Hosting options for urban deployment of virtualized small cell function .11

Figure 5-1

Example PNF/VNF management co-ordination via OSS/BSS ................14

Figure 5-2

Services integration using a virtualized ESCG/ESCC that is i ntegrated with the centralized small cell virtualized function VNF. ......................15

Figure 5-3

Analysis estimating the CDF of transport delay associated with existing enterprise WAN service and mapping to fronthaul types..........15

Figure 5-4

Peak-to-mean non-uniform spatial distribution ..................................16

Figure 6-1

TCO comparison of different splits ....................................................18

Figure 8-1

Summary of benefits versus costs of various different split options ......22


1. Introduction The virtualization of network functions is an industry transition th at is impacting all service provider segments and will therefore effect the realization of future networks of mobile network operators. Previous SCF publications have focused on the use of virtualization of small cell core network functions [ 6], translating concepts pioneered by ETSI NFV and applying those to the Small Cell Core Network domain, representing ETSI use case #5 (Virtualization of Mobile Core Network and IMS) [ 7]. Compared with this previous analysis, this document describes the output of a study by the Small Cell Forum to address ETSI use case #6 (Virtualization of Mobile Base Station) as applied to small cells. From a RAN perspective, virtualization and centralization may be viewed as opposing conventional macro evolution. The shift from 2G, through 3G to 4G has seen increasing functionality shift into the base station and a corresponding flattening of the access architecture. The small cell industry has leveraged this transition, enabling innovative capabilities to be delivered with small cell access points that offer hotspot capacity and coverage. Contrary to this evolution is a shift towards virtualization and cloud. Driven by transitions in the datacenter, workloads are in creasingly being centralized. This transition is impacting the wireless world wi th the advent of cloud RAN (C-RAN). In its purest form, this is the opposite philosophy to convectional RAN evolution (and small cell definition), with total centralization: having a ‘dumb’ radio at the edge (RRH or Remote Radio Head) with all the signals sent back over fiber (‘front haul’ - as opposed to the backhaul out of a base station) to a server farm that does all of signal processing for the whole network.

Figure 1-1

Contrasting evolutions of the RAN architecture

These contrasting evolutions, illustrated in Figure 1-1, have been used to position small cells and virtualization as an ‘either or’ ar chitectural approach. Importantly, cloud RAN facilitates the support of advanced signal processing techniques defined in LTE-A that rely on tighter coordination between base stations. Techniques such as CoMP, macro-diversity or 3D MIMO rely on fast, l ow-level communication between different sites; that is complex to realize with a decentralized flat RAN architecture but trivial with a centralized approach. But Cloud RAN typically requires ideal transport/dark-fiber support that is likely cost-prohibitive for supporting cloud RAN capabilities in a small cell environment that has been defined to be able to be transported over consumer grade broadband networks. Consequently, whilst the current description of ETSI use case #6 has concentrated on a macro-centric view of base station functions, describing techniques to leverage the Report title: Virtualization for small cells: Overview Issue date: 09 June 2015 Version: 106.07.01

1

common public radio interface (CPRI) base station decomposition that requires ideal fiber transport (see www.cpri.info), the Small Cell Forum is motivated t o studying alternative approaches to decomposing small cell functionality and in particular those approaches that are suitable for being transported over the packet switched transport networks (with their associated bandwidth and delay/jitter characteristics) conventionally used to support small cell deployments. In particular, this small cell decomposition that facilitates network function centralization and/or virtualization, can enable a range of new capabilities by the small cell layer, including: •

•

•

•

•

•

•

•

•

•

•

Enabling the definition of one single virtual cell that is supporting multiple physical remote small cells Supporting scalable hierarchical mobility whereby inter-small remote small cell mobility is hidden from upper layer elements Facilitating policy enforcement to be applied at an aggregate level, e.g., enabling admission control type capabilities Enhancing the security of the small cell solution by terminating user-plane encryption above the remote small cell Enabling statistical multiplexing of compute resources which lowers peak-tomean ratio of the load experienced by the centralized function. This can then be leveraged to lower footprint and/or energy consumption of the system Facilitating the deployment of advanced radio techniques, e.g., CoMP (including coordinated scheduling and beamforming), carrier aggregation (including cross carrier scheduling), high order MIMO, to enhance the coverage and/or capacity of the virtualized small cell system Supporting enhanced SON operation by providing visibility of operation across a cluster of physical remote small cell units, including allowing dynamic resource allocation and traffic load balancing Improved future proofing by being able to add additional processing at small number of accessible central locations in contrast to upgrading a large number of individual small cells in less accessible public spaces. Simplifying the remote management of the man y physical network functions as capability is relocated into the centralized VNF component Leveraging standard NFV Infrastructure by moving at least part of the basestation on to standard IT servers, storage and switches Enabling NFV based service-chaining, e.g., as described in [SCF154], to integrate the virtualized small cell with other VNF-based functions, including creating a competitive environment for innovative third party applications by unlocking proprietary boundaries

Examining the above capabilities, it is evident why the Small Cell Forum’s study has primarily focused on the issues associated with decomposing th e small cell (or more generally the base station) protocol stack and its segmentation between a physical network function (PNF) and a centralized element that is then suitable to be realized as a virtual network function (VNF). However, one of the aims of this small cell virtualization study has been to take a more holistic view of the topic, looking at technology, system and business implications of applying virtualization techniques to the small cell radio access network and so thi s document also examines the system architecture issues with supporting a virtualized small cell. Note, while the main focus of this study is small cells, it is expected that some of the results may be applicable to the virtualization of base stations of all sizes.


2

1.1

Structure of this document

This overview is structured around the different disciplines represented by the Forum’s working groups. Compared with the detailed definition work of previous releases that involved the entire Forum’s working groups, the virtualization study represents a more focused analysis of the topic. The areas covered i nclude: •

•

•

•

•

1.2

Market drivers: Examines the business case impacts from a TCO perspective associated with adopting small cell virtualization Motivations for small cell virtualization study: Describes the rationale behind the study from the Forum’s Operator Group Small cell decomposition: Looks at alternative approaches to decomposing small cell functions between physical and virtual components Coverage, capacity and performance aspects of small cell virtualization: Examines the impact of virtualization on being able to support various advanced multi-point co-ordination capabilities to deliver capacity and/or coverage capabilities Network aspects of small cell virtualization: analyses the impact of base station virtualization on the end-to-end small cell system

Value of SC-virtualization content for different audiences

Operators can use the study to develop their own access network virtualization strategy and in particular will be able to frame out the issue of RAN virtualization from a classical macro-CPRI perspective and contrast that with alternative virtualization approaches suitable for being transport of networks t hat cannot support CPRI based decomposition requirements. Vendors can better understand the opportunity to evolve their small cell architectures to address increasing levels of virtualization. It will help to refine their roadmaps to ensure they are able to leverage the latest virtualization concepts whilst being aligned with industry transition to virtualization. Regulators can understand the impact of RAN virtualization and multi-tenancy on regulatory aspects and have a better understandi ng of the decomposition of base stations into Physical Network Function and Virtual Network Function. Partner organisations engaged in network function virtualization are able to understand how the Forum is positioned to address ETSI NFV Use Case #6 related to base station virtualization, understand possible alig nment of virtualization requirements across different RAN types and possibly leverage cross-organizational synergies.

1.3

Terminology

There is a general lack of consensus in the industry as to the t erminology used to describe the different aspects of RAN virtualization. For example, even though ‘C-RAN’ is a popular term, there is ambiguity around whether the ‘C’ refers to centralized, coordinated or cloud. Whilst there is less ambiguity in ETSI NFV as to the characteristics of a virtual network function (an implementation of a network function that can be deployed on a network function virtualization infrastructure) compared to a physical network function (an implementation of a network function via a tightly coupled software and hardware system), such terminology is not small cell, or RAN, specific. Hence, the following terms are defined to help describe small cell virtualization: Report title: Virtualization for small cells: Overview Issue date: 09 June 2015 Version: 106.07.01

3

Virtualized small cell: The applicability of NFV techniques to small cell base stations whereby a subset of small cell functionality is run as one or more VNFs on virtualized compute platforms and the remaining functionality is run on a physical network function. The analysis of the different split options b etween virtualized and physical network functions is one of the key considerations of th e Small Cell Forum’s study. Centralized small cell: The virtualized small cell capability where the compute platform hosting the Network Function comprising the subset of small cell functions is remote from the remaining non-virtualized functions and is implemented as a VNF. Remote small cell: The network functionality of the virtualized small cell that encompasses the RF termination point and associated non-virtualized functions that is implemented as a PNF. The combination of centralized small cell and remote small cell provides all conventional small cell functionality. Small cell fronthaul: The transport system used to provide connectivity between the centralized small cell and the remote small cell. The Fronthaul transport system will have associated bandwidth and delay/jitter characteristics. Grouping of characteristics may be defined, e.g., describing fronthaul transport as ‘ideal’, ‘near ideal’, ‘non-ideal’, etc. Small cell resource coordination: The evolution of the centralized small cell, whereby the centralized functionality includes co-ordination between functions controlling multiple remote small cells. This coordination may provi de enhanced functionality compared to standardized X2 based co-ordination approaches.


4

2. Motivation for small cell virtualization study In June 2014, the Forum’s Operator Group met to discuss the impact of RAN virtualization on the adoption of small cell based architectures. In particular, from an industry adoption perspective, it was clear that cloud RAN and small cell deployment were no longer being seen as an ‘either/or’ approach, with research from MaravedisRethink indicating that the majority of operators deploying small cells will also look to deploy virtualization in parts of their RAN and that over 90% of those operators deploying cloud-RAN will also be using urban small cells, as illustrated in Figure 2-1.

Figure 2-1

Analysis of market adoption of RAN virtualization and small cells

However, whilst the above figure indicates that cloud-RAN and small cells should no longer be presented as alternatives, there are issues concerning the evolution of the radio access network that need to be addressed. For example, how should clusters of small cells best interwork with a virtualized macro layer? Furthe rmore, having competing approaches to virtualization in the macro and small cell may inhibit cooperation between the layers and reduce the operator’s return on investment. Indeed, when polling operators on the key barriers to adopting virtualization in the access network, a survey by Maravedis-Rethink indicated that a l ack of a common approach to virtualization would ultimately hinder adoption, as illustrated in Figure 2-2.


5

Figure 2-2

Factors identified by MNOs as the key barriers to adopting virtualization in the access network

As the mission of the Small Cell Forum is to reduce the barriers to adoption of small cell technologies across a range of use cases, it was clearly evident that the issue of virtualization was within the remit of the forum. However, unlike other bodies of work from the Forum which were focused on the adoption of small cell technology by particular well defined market segments, this virtualization issue was not r estricted to a single market segment. As a consequence, the SCF’s board concluded that the subject of small cell virtualization merited a broad analysis and study. To frame out this small cell virtualization study, the forum’s Operator Group formulated key questions that needed answering concerning the adoption of virtualization technologies in the access network: 1. Is there any divergence between how virtualization is viewed from a macro RAN perspective versus a small cell perspective? Background: Importantly from a deployment perspective, small cells are predominantly transported using IP/Ethernet based backhaul networks. Thi s can be contrasted with the initial proposals to virtualize the macro n etwork which are based on CPRI and an optical fiber based transport network with very tight tolerances on delays. 2. What are the real business benefits of virtualization in the RAN? Do benefits apply equally to macro and small cell deployment use cases? Background: Some of the claimed advantages of virtualizing the macro layer have been around total cost of ownership and benefits of deploying small form factor Remote Radio Heads (RRHs) compared with composed traditional base station equipment. Small cells are already small form factor and hence some of the claimed TCO benefits may not equally apply to both macro and small cell.


6

3. Are current small cell architectures well suited to virtualization/cloudification? Background: Whereas conventional macro based cloud RAN prop ositions are based on a CPRI decomposition with ideal transport ch aracteristics, small cell architectures have been developed to address deployment scenarios that include residential use cases where transport utilizes consumer grade broadband se rvices. Can a small cell proposition that has been specifically architected to be supported using non-ideal backhaul networks be adapted to support virtualization and/or cloudification? 4. Are there unique attributes of small cell transport that impact fronthaul considerations? Background: Previous questions have highlighted the key role of transport characteristics for supporting virtualization in the access network. Key aspects such as synchronization, sharing as well as latency and bandwidth parameters may impact the ability to support virtualization scenarios. 5. Can virtualization deliver a phased roadmap – e.g., aligned with longerterm 5G directions? Background: This observation recognizes that virtualization is being applied retrospectively to existing small cell architectures. However, moving forward, it is anticipated that as 5G gets defined, RAN virtualization is accepted as a core requirement that needs to be accommodated by the 5G access architecture. However, being able to apply virtualization to existing 4G access network can enable operators to invest in capabilities that can then be re-used as operators eventually migrate towards 5G based access networks. 6. How do we ensure virtualization supports innovation across a multi-vendor ecosystem? Background: One of the primary successes of the Small Cell Forum has been the enablement of the delivery of multi-vendor small cell systems. Compared with these standardized approaches, the small cell industry is already showcasing examples of applying virtualization techniques to the small cell layer. However, the se are currently exclusively based on proprietary approaches. Hence, a key qu estion to consider is whether virtualization and multi-vendor are mutually exclusive approaches or whether the successful multi-vendor ecosystem can be evolved to address new virtualization opportunities.


7

Figure 2-3

Complex issue required to be addressed when studying small cell virtualization

These questions from the Operator Group triggered the initialization of the Forum’s virtualization study. Figure 2-3 highlights how the complexity of the virtualization issue, requiring a set of competencies that are well matched to those of the Forum: •

•

•

•

•

In particular, analysis of the issue requires capabilities for examinin g the business case drivers for RAN virtualization, enabling the forum to leverage its competencies in small cell business case analysis. The overall business case will be influenced by the coverage and capacity characteristics that can be beneficially supported by the vi rtualized small cell layer, a set of skills that the Forum has in its RPH working group. Virtualization requires additional deployment of the virtualized small cell component and existing SCF architecture work can be leveraged to understand the impact of virtual network function deployment on the individual virtualized small cell use cases. The SCF has expertise in analysing the challenges in supporting small cell backhaul within its Backhaul/Transport group. These same transport competencies can be re-applied to analyse the transport of the new fronthaul interfaces. Finally, the virtualization of the small cell layer has impacts on the end-toend small cell system, expertise that exists within the SCF’s NET Working Group.


8

3. Small cell decomposition: virtualization use cases Whilst extensive work on network function virtualization (NFV) has been performed by ETSI, the focus to date on ‘Use case #6: Virtualization of the Mobile Base Station’ has been predominantly around the macro RAN use case. Examining use case #6 in more detail highlights a range of capabilities described in section 1 that motivate the drive to virtualize the base-station. These drivers are equally applicable to small cell, including residential small cells, enterprise small cells, urban small cells and rural and remote small cells. However, unlike other virtualization use cases, applying NFV techniques to a base station still necessitates a physical network function (PNF) that is responsible at least for supporting the RF functions of the base station. This necessity to decompose a small cell into physical and virtual network functions is the focus of [1], ‘Small cell virtualization use cases’. Compared with the baseline macro use case o f virtualization that is based on CPRI, [1] analyses a range of different decomposition options as illustrated in Figure 3-1.

Figure 3-1

Various possible LTE base station decompositions

The rationale for examining these alternative splits is related to the associated requirements on the transport network for supporting the fronthaul link between the VNF and PNF components. As an increasing set of functions are implemented as a virtual network function, the transport requirements in terms of bandwidth and latency become more onerous. In particular, the conventional macro-cell decomposition relying on CPRI requires bandwidth in excess of 2.5Gbps for supporting a standardized LTE 2x2 20 MHz base station that today may be backhauled using 150 Mbps of bandwidth for user and control traffic. More critically, the latency requirements of conventional backhaul transport are typically bounded by the application requirements, for example with 3GPP recommending an upper bound of 30 ms for supporting real -time gaming services. With the same CPRI based decomposition requiring transport latencies less than 250 µs, it is clear that fronthaul transport characteristics have a significant impact on the types of decomposition that can be supported.


9

[1] analyses different alternatives for functional decomposition between PNF and VNF, including splits corresponding to the classical layering (PDCP/RLC/MAC/PHY) within an LTE small cell, as well as ‘fractional splits’, for example ‘Split -MAC’ and different ‘SplitPHY’ approaches. Table 3-1 below summarizes the latency and bandwidth characteristics associated with these decomposition use cases (detailed assumptions are included in [1]). Note: [2] also addresses the operation of the virtualized small cell when the transport bandwidth is below the listed values. Use Case

One-Way Latency

DL Bandwidth

UL Bandwidth

PDCP-RLC

Non Ideal – 30ms

151Mbps

48Mbps

RLC-MAC

Sub Ideal – 6ms

151Mbps

48Mbps

Split MAC

Sub Ideal – 6ms

151Mbps

49Mbps

MAC-PHY

Ideal – 250µs Near Ideal – 2ms

152Mbps

49Mbps

PHY Split I


173Mbps

452Mbps

PHY Split II


933Mbps

903Mbps

PHY Split III


1075Mbps

922Mbps

PHY Split IIIb


1966Mbps

1966Mbps

PHY Split IV

Ideal – 250µs

2457.6Mbps

2457.6Mbps

Table 3-1

Latency and bandwidth characteristics associated with these decomposition use cases.

In terms of deployment options, the definition of the virtualized small cell function raises questions about where to locate such capability. [1] examines the use of virtualization within the broad range of small cell deployments, highlighting a range of options for hosting such functionality. The hosting options for a large enterprise deployment are shown in Figure 3-2, illustrating that the virtualized small cell function can be hosted either on site, by the enterprise’s WAN service provider, geographically located in an MNO’s regional data center or at the perimeter of the MNO’s network.

Figure 3-2

Hosting options for large enterprise deployment of virtualized small cell function

Urban deployment of the virtualized small cell function enables different connection opportunities for the remote Physical Network Function, including Report title: Virtualization for small cells: Overview Issue date: 09 June 2015 Version: 106.07.01

10

•

•

•

Macro – PNF is chained from a macro site with connection into the operator managed network Network termination point – specific edge termination point into the operator managed network Metro Ethernet – metro owned and managed backhaul

Accordingly, hosting options for the virtualized small cell function in the urban use case are illustrated in Figure 3-3 and include: • • • • • •

On macro site Metro network data center Operator data center Metro service provider Geographically locally located or regional data center Operator’s edge service provider

Figure 3-3

Hosting options for urban deployment of virtualized small cell function

A key conclusion of the use case analysis in [1] is that due to the wide and ranging possible small cell deployment scenarios, one size fits all – that is a virtualization approach based on a single decomposition that can b e applied to all deployment scenarios - is unlikely to be able to cover all scenarios. To help characterize the different options, the generic fronthaul transport network is characterized as supporting one of four different capabilities: Fronthaul option

Transport characteristics

Non-ideal

Supporting a one-way latency of up to 30ms, <10ms jitter, limited and variable bandwidth

Sub-ideal

Supporting a one-way latency of up to 6ms, <2ms jitter and un-constrained bandwidth

Near-ideal

Supporting a one-way latency of up to 2ms, with minimal jitter and bandwidth in excess of 2.5 Gbps (for a 20 MHz 2x2 small cell)

Ideal

Supporting a one-way latency of less than 250µs, with minimal jitter and bandwidth in excess of 2.5 Gbps (for a 20 MHz 2x2 small cell)

Table 3-2

Fronthaul transport options


11

4. Coverage, capacity and performance aspects of small cell virtualization Building on the use case analysis, four particular decompositions are an alyzed in detail around the benefits that can be achieved by deploying a centralized (virtualized) small cell function. Such a centralized capability can be used to coordinate resources among a cluster of remote small cells/physical network functions that may affect the performance of the small cell network from a capacity and coverage perspective. [2] describes how the centralization of functions that is enabled through virtualization enables more advanced physical layer performance enhancement techniques, typically found on large macrocells, to be used in small cell deployments. These techniques include the coordination scheduling and the processing of information between multiple cells and multiple antenna within a cell, which, to date, have not normally been available on small cell systems. Exactly which techniques can be used depends on the functional decomposition between PNF and VNF components, which in turns also depends on the performance and capabilities of the fronthaul link between them. The techniques described in SCF160 include: • • • •

Carrier aggregation (inter and intra band) Cross-carrier scheduling High Order MIMO DL and UL CoMP , including beamforming/coordinated scheduling

[2] includes a comprehensive review of 3GPP and NGMN analysis associated with CoMP performance when operating over i deal and non-ideal backhaul. Whereas ideally the analysis into coverage and capacity would be able to provide quantitative analysis of the enhanced coverage and/or capacity benefits of small cell co-rdination, the experience of 3GPP in simulating coordinated multi -point performance has highlighted that, even with very well defined assumption, there was significant divergence in reported results. Hence, whilst there w as general consensus that coordinated multipoint functionality would normally improve performance, the lack of agreement in 3GPP on the degree of improvement had consequences on the associated small cell analysis. [2] analyses four decompositions in detail, namely PDCP/RLC, Split-MAC, MAC/PHY and Split-PHY. Table 4-1 shows these splits and the various different CoMP techniques that are applicable to the different split s. The key takeaway is that the lower d own in the protocol stack the decomposition occurs, the greater the ability to benefit from the enhanced co-ordination techniques. For example, a decomposition based on PHY Spl it III supports all possible techniques, whereas a decomposition based on PDCP/RLC only supports co-ordinated scheduling/co-ordinated beamforming techniques.


12

Advanced RF combining capability

PDCP/ RLC

Split MAC

MAC/ PHY

Split PHY

Carrier aggregation

✔

✔

✔

Cross carrier scheduling

✔

✔

✔

High order MIMO

✔

Down link joint processing –joint transmission (JT)

✔

✔

Uplink joint reception (JR) independent PHY decoding

✔

✔

Uplink joint reception (JR) joint equalization PHY decoding

✔

Join processing – dynamic point selection (DPS) Coordinated scheduling/beamforming (CS/CB) UL and DL Table 4-1

✔

✔

✔

✔

✔

✔

✔

Mapping of enhanced co-ordination techniques to virtualization use cases splits

One of the challenges introduced with operating a decomposed small cell over a fronthaul link is associated with the additional latency and in p articular the limited delay budget necessary to meet HARQ processing requirements. [2] discusses the use of HARQ interleaving that uses standardized signaling to defer buffer emptying, enabling higher latency fronthaul links to be accommodated. Such a technique is applicable to PHY Split III, MAC-PHY and Spli t-MAC based decompositions. While HARQ interleaving functionally is standardized, its operation does add addi tional latency to the overall transmission flow in both the downlink and uplink, and may, when a decomposed small cell is serving only a limited number of UEs, limit the peak cell throughput. Note that, while increasing the number of UEs active in a cell will mitigate this impairment from an aggregate cell capacity perspective, a single UE will never be able to achieve peak cell throughput (e.g., as represented by a single user in a cell).


13

5. Impact of base station virtualization on the end-to-end small cell system Whereas [1] and [2] are primarily focused on issues associated with the base station protocol stack and its segmentation between a p hysical network function (PNF) and a centralized element that can then be realized as a virtual network function (VNF), one of the aims of this small cell virtualization study is to understand the impact of base station virtualization on the end-to-end small cell system. [4] examines the impact of small cell virtualization on the currently defined management architecture. Using ETSI-NFV and 3GPP architectures that support combined PNF and VNF systems, this document frames the evolution of the current multi-vendor small cell management system to address this functional decomposition. Figure 5-1 shows one example of how co-ordination can be realized between PNF and VNF components. As the 3GPP Small Cell management architecture leverages many of the concepts originally defined by the Broadband Forum (BBF), [4] examines how virtualization is impacting BBF defined architectures, drawing parallels between the forum’s small cell decomposition study and on-going BBF work defining a Network Enhanced Residential Gateway (NERG), where a NERG is defined as a ‘virtual CPE’ where functions of the CPE reside both in the subscriber’s premises and in the network service provider’s network.

Figure 5-1

Example PNF/VNF management co-ordination via OSS/BSS

From an enterprise small cell architecture perspective, attention i s focused on the impact of virtualization on the previously defined enterprise small cell network architecture, highlighting similarities between the centralized small cell virtualized function and previously defined Enterprise Small Cell Concentrator functionality. Network Function forwarding graph techniques are described that enables the decomposed small cell to be integrated with in existing enterprise architectures, as illustrated in Figure 5-2.


14

Figure 5-2

Services integration using a virtualized ESCG/ESCC that is integrated with the centralized small cell virtualized function VNF.

Examining the enterprise use case in more detail, a survey of WAN bandwidths together with typical configuration was used to estimate the quality of Enterprise WAN service to support different fronthaul characteristics. Figure 5-3 summarizes the analysis, indicating that 35% of WAN services were estimated to have characteristics able to support non-ideal fronthaul. More p recisely: •

•

•

20% of the total, or 57% (20/35) of the ‘fronthaul compatible’ WAN services, were able to support sub-ideal fronthaul 10% of the total, or 28% (10/35) of the ‘fronthaul compatible’ WAN services, were able to support near-ideal fronthaul, and 2% of the total, or 5% (2/35) of the ‘fronthaul compatible’ WAN services, were able to support ideal fronthaul service.

Figure 5-3

Analysis estimating the CDF of transport delay associated with existing enterprise WAN service and mapping to fronthaul types


15

Finally, the benefits of virtualization/centralization as it relates to efficient resource utilization are analyzed, with techniques described that can leverage th e non-uniform spatial and temporal utilization to improve the efficiency of virtualized small cell deployments. Figure 5-4 shows one example of non-uniform spatial distribution t hat can be leveraged by a more centralized VNF, in this example illustrating a 4-to-1 gain in terms of resource allocation compared with a conventional distributed small cell approach.

Figure 5-4

Peak-to-mean non-uniform spatial distribution


16

6. Examining the business case impacts for small cell virtualization The use case analysis in [1] indicates the importance of fronthaul characteristics when considering the options available for small cell virtualization. However, the f ronthaul transport has a significant impact on the overall virtualized small cell business case. Key to the business case drivers for virtualized small cells is to understand the costs associated with higher performance fronthaul transport versus the benefits in terms of improved RF performance (covered in [2]) and/or improved resource utilization (covered in [4]). While the Forum has previously avoided pub lications examining direct cost comparisons between backhaul solutions based on different technologies, the impact of transport on the overall TCO of virtualized RAN was felt significantly important to merit analysis. Instead of publishing results of an in ternal investigation, the Forum contracted Real Wireless to produce an independent analysis. [3] describes the results of this Real Wireless study into the cost elements for a range of small cell RAN virtualization options. Whereas the Forum is interested in analyzing the business benefits of virtualization in all deployment use cases, the dominance of transport in the overall analysis focused the study on those deployment use cases where transport costs were well defined. This led to the analysis being focused on the urban deployment of virtualized small cells, compared with enterprise use cases where transport can be provided by the enterprise LAN/WAN and residential that leverages incumbent consumer broadband connectivity, both at very low marginal costs. Leveraging earlier work by the Forum examining the business case for urban small cells [8], the analysis investigates where a virtualized small cell network architecture differs from that of a traditional small cells in cluded, specifically looking at impacts on: •

• •

Transport costs for a range of transport technologies and the throughput and latency requirements of the various RAN d ecomposition options; Deployment costs; Power consumption (driving electricity costs) that varies according to the change of processing in the core (virtualised functions of core servers) versus in the small cells.

Leveraging feedback from a wide range of stakeholders, including five wireline transport providers, six wireless transport providers, five mobile operators and one silicon vendor, the analysis examines a number of different transport options available: •

•

•

•

•

•

Managed fiber: where the capacity can be leased from the fiber service providers using their existing fiber network. Dark fiber: this is a fiber connection that is not ‘lit’. The fiber provider will typically splice an extension to an existing fiber in order to terminate the connection but the end user has the responsibility to installing the terminating equipment. Copper (VDSL2 and G.fast): this is very high bit rate digital subscriber line, usually provided by the leased line providers. Sub-6 GHz: wireless communications systems that uses sub 6 GHz licensed and unlicensed spectrum. Microwave (6 - 42 GHz): wireless communications systems in point-point and Point–to-multipoint forms. mmWave (60 GHz): wireless communications systems that provide pointto-point communications using 60 to 80 GHz bands.


17

For each transport option, the support for the various virtualization splits defined in [1] is discussed together with future evolutions of the particular transport system to enable possibly unsupported splits to be accommodated in future years. Importantly, cost aspects associated with both capital and operational aspects are combined into an overall TCO figure.

5 year TCO (in 2020) of different splits 100 90 80 70 60 k 50 $ 40 30 20 10 0

Figure 6-1

Dark Fibre Managed Fiber Wireless (60GHz) Copper S1 (D-RAN)

PDCP/RLC

MAC/PHY

CPRI,ORI

188Mbps

188Mbps

188Mbps

2.5Gbps

50ms

30ms

2-6ms

250us

TCO comparison of different splits

Source: Real Wireless

Figure 6-1 shows the variation in TCO for different functional splits with different transport technologies. In most cases it shows the hi gher performance required by different functional splits has little impact on the TCO. In the managed fiber case, the CPRI split would be charged at a premium due to the tougher SLA. Copper is not expected to evolve to support the CPRI split. The analysis also indicates significant regional differences in the TCO comparisons between services available in Europe, US and Japan.


18

7. Virtualization study conclusions Given that this study into virtualization was initiated by a set of questions posed by the Forum’s Operator Group, it is instructive to examine whether the study has been able to answer the particular issue. Q1. Is there any divergence between how virtualization is viewed from a macro RAN perspective versus a small cell perspective? The study concluded that in contrast to the ideal transport required for supporting macro-cell decompositions based on CPRI/ORI, a range of transport characteristics can be supported for virtual small cells, including those that may be able to be supported within conventional small cell deployments in residential, enterprise and urban environments. Q2. What are the real business benefits of virtualization in the RAN? Do benefits apply equally to macro and small cell deployment use cases? Virtualization gains associated with the non-uniform spatial and temporal consumption of mobile data and the resource efficiencies obtained by centralizing functionality are equally applicable to macro and small-cell virtualization. Gains in terms of radio efficiencies are associated with the selected decomposition, with an increasing set of capabilities able to be applied the further down the radio stack the split is performed. When virtualization is performed using a PDCP/RLC and MAC/PHY splits, virtualization can be realized with no increased transport costs and is able to be supported over all technologies. Conversely, virtualization based on the classical CPRI/ORI split requires low latency high bandwidth services that are currently unable to be supported over managed fiber and copper based transport services. Q3. Are current small cell architectures well suited to virtualization/ cloudification? Previous SCF architectures have highlighted the benefits of hierarchical mobility scaling with the introduction of the enterprise small cell concentrator [9] that can be seen as providing some of the benefits of centralization as described in the introduction. Further, in [6] the opportunities associated with applying Network Function Virtualization to the small cell core network have been described including these hierarchical concentrator functions. Hence, applying virtualization to the small cell itself can be seen as an evolution of already defined enterprise small cell architectures, enabling the benefits of centralization and virtualization to be appli ed across all small cell market segments Q4. Are there unique attributes of small cell transport that impact fronthaul considerations? Much of the focus of the small cell virtualization study has been associated with the impact that the transport characteristics have on the ability to decompose a small cell into a physical network function and a virtual network function. Compared with traditional techniques that require ideal-transport for fronthaul connectivity, i.e. dark fiber, the study has examined alternative decompositions that can be supported over a range of transport options, enabling strict requirements on fronthaul latency and bandwidth to be reduced.


19

Q5. Can virtualization deliver a phased roadmap – e.g., aligned with longerterm 5G directions? This study has highlighted that applying network function vi rtualization to the radio access network is a trend that will likely be adopted in advance of any 5G definition. Whilst applying virtualization retrospectively to a pre-existing RAN architecture is indeed more complex than including virtualization as a foundational requirement when a new RAN architecture gets defined, the study has highlighted those key lower layer RAN functionalities and transport characteristics that impact the supported decomposition and associated virtualization approach. Decomposing an LTE base station into a PNF and VNF component in advance of 5G will surely provide those operators deploying such an architecture valuable competencies and insights that can be applied to any future 5G deployments. Q6. How do we ensure virtualization supports innovation across a multivendor ecosystem? The small cell ecosystem has already embraced FAPI based base station decomposition, based on a MAC/PHY split [ 10], which enables small cell manufacturers to adopt a multi-vendor silicon strategy. Further, the Sm all Cell Forum, in co-operation with ETSI and NGMN alliance have been working together since 2010 to conduct multi-vendor Plugfests to accelerate the alignment of small cell network technologies [11]. These two proof points highlight that multi-vendor virtualized small cells based on an agreed decomposition is an achievable goal that the i ndustry can aspire towards.


20

8. Summary Virtualization is set to impact conventional small cell architectures, but its adoption requires a balancing act to be performed between often conflicting requirements of a need to support the widest possible range of transport characteristics, the need to deliver timely multi-vendor systems, the opportunity to enhance RF performance through the use of multi-point co-ordination and the prospect of being able to leverage COTS datacenter hardware to host the VNF components. Indeed many of the claimed benefits of virtualization are in fact due to the centralization of functions realized by a decomposition of a conventional small cell into two functional elements. These include: •

• •

•

• •

Enhanced scalability whereby (frequent) mobility events are masked from upper layers and new cells can be easily added to the network Improved inter-cell interference co-ordination and SON capability Reduced security exposure by re-locating air-interface security termination from the small cell to a more central location Facilitating policy to be applied to aggregate level – e.g., WAN based admission control in the enterprise Simplifying PNF management compared with full small cell Avoiding complication of deploying intercept points at the small cell/PNF location when delivering enhanced services

The benefits associated the virtualization of these central functions include: •

• •

•

•

Reduction in resources because of lower p eak-to-mean of aggregated traffic/events/users Easy to scale up centralized component with standard compute Facilitates upgrade lifecycle: new features can be more rapidl y deployed on a virtual platform – and wound back if needed. Flexibility: enables functions to be moved around depending on loading conditions or availability of compute and transport capacity resources. Standard NFV service chains to couple small cells to other VNFs, including data plane handling functions as per SVC/MEC definitions

However, such decomposition into centralized and remote functionality comes at the cost of tighter constraints on the transport characteristics, the more base station functionality that is centralized, the more onerous these requirements. Today, two extremes are available: •

•

Classical small cell: whereby all base station functionality is distributed to the physical network function and backhaul requirements accommodate low cost consumer grade broadband CPRI based C-RAN: whereby all base station functionality with the exception of RF is centralized and fronthaul requirements may require expensive dark fiber deployments

This study has demonstrated that when compared with these extremes, there is a range of small cell base station decompositions that are feasible, enabling operators to benefit from centralizing functionality while being able to trade off the transport requirements versus radio performance of the alternative approaches, and enabling virtualization to be applied to the small cell radio access network, as illustrated in Figure 8-1.


21

Figure 8-1

Summary of benefits versus costs of various different split options

Specifically, the analysis concludes that the MAC/PHY split delivers most of the benefits of centralization, with only a small increase in transport performance and is well aligned with the current small cell multi-vendor ecosystem approach based on the functional application platform interface (FAPI). For those use cases that face restrictions from a transport perspective such that the backhaul system cannot be enhanced to support the MAC/PHY split, then the PDCP/RLC split has the potential to deliver a subset of the centralization benefits b ut with no additional transport requirements when compared with conventional small cell deployments. Given such a positive analysis, the Forum is motivated to define a transportable interface for the spit small cell, to establish a scalable ecosystem with a converged approach to virtualization across multiple suppliers.


22

References 1 2 3 4 5 6 7 8 9 10 11

[SCF159] ‘Business case elements for small cell virtualization’, Small Cell Forum [SCF160] ‘Coverage and capacity impacts of virtualization’, Small Cell Forum [SCF158] ‘Business case elements for small cell virtualization’, Small Cell Forum [SCF161] ‘Network aspects of virtualized small cells’, Small Cell Forum [SCF106] ‘Overview of virtualization for small cells’, Small Cell Forum [SCF154] ‘Case histories for rural and remote small cells’, Small Cell Forum ‘Network Functions Virtualization (NFV): Use Cases’, ETSI GS NFV 001 [SCF087] ‘Business case for urban small cells’, Small Cell Forum [SCF067] ‘Enterprise small cell network architectures’, Small Cell Forum [SCF082] ‘LTE eNB L1 API definition’, Small Cell Forum [SCF085] ‘Value of Small Cell Forum Plugfests’, Small Cell Forum


23

Overview of Virtualization for Small Cells

Recommend Documents