creating cloud base stations with ti s keystone … called the cloud radio access network (c-ran),...

Overview

The Cloud RAN (Radio Access Network) is a

hot topic in the wireless community. Cloud

RAN takes its place along with small cells,

distributed antenna systems (DAS), ac-

tive antenna arrays (AAA) and remote radio

heads (RRH) that are being considered, or

in some cases deployed, to address wire-

less issues ranging from site acquisition,

to coverage and capacity enhancement, to

environmental stewardship (green base sta-

tions). This paper explores how enhanced

versions of TI’s KeyStone Multicore Architec-

ture can be used to create super-sized cloud

base stations. All of the concepts discussed

exist in the KeyStone architecture today*.

Creating cloud base stations with TI’s KeyStone multicore architecture

BackgroundThe traditional wireless network topology consisting of adjacent or overlapping macro cells

is the workhorse of our industry. Looking ahead, various factors are driving change to this

topology, including capacity constraints, coverage issues and a need to support new classes

of users and devices. Smaller specialized cells for indoor use, hot spots, and focused outdoor

use are beginning to be deployed. In parallel, semiconductor and architectural innovations are

enabling chips with performance and functional integration that was unimaginable a few short

years ago. These factors have led Texas Instruments (TI) on a venture to design System-on-

Chips (SoCs) targeting complete base stations on a chip (Figure 1). These SoCs are based on

TI’s innovative KeyStone multicore architecture. Against this backdrop, the prospect of a base

station using cloud computing concepts is being contemplated. In the proposed implementa-

tion, called the Cloud Radio Access Network (C-RAN), the functions of a base station are bro-

ken apart with some in the field some in a central site. The reality is that each of these base

station formats, macro cell, small cell and cloud based will fill specific niches and coexist in

future networks. Each format will benefit from innovations in TI’s KeyStone multicore architec-

ture as it is currently deployed and from future enhancements that are in development today.

Tom Flanagan,Director, Technical Strategy

Wireless Base Station InfrastructureTexas Instruments

W H I T E P A P E R

Figure 1. The migration to base station on a chip.

* Capacity enhancements to existing KeyStone in-frastructure features, some of which are currently in development, are required to fully implement very large scale device-to-device interconnectiv-ity with the devices acting as a single logical unit.

Creating cloud base stations with TI’s KeyStone multicore architecture October 2011

2 Texas Instruments

In traditional base stations, the digital and radio processing elements are co-located with the antenna

array. This is true for both macro and small cell formats, though there are variations on the implementation

scheme. For example, remote radio heads are now relatively common in macro deployments. In this configu-

ration, the radio equipment is located with the antenna array and the baseband digital processing, called the

Radio Equipment Control, is remote with a fiber link connecting the two. Picture the radio elements being at

the top of the tower and the digital processing in the cabinet at the base of the tower. The Common Public

Radio Interface (CPRI) was developed to facilitate this scenario. It provides a standard interface for intercon-

necting the radio elements to the baseband processing so that the operator can mix and match these com-

ponents from different vendors (although they rarely do). The major difference between the classic remote

radio head and the cloud base station is the length of the fiber. In the first instance the fiber is relatively short,

a few hundred feet at most. In a cloud topology the fiber run may be many kilometers long. In addition, there

is extra processing deployed at the antenna to compress the antenna data so that multiple antenna streams

can access and share the long-haul fiber.

The benefit of a cloud base station, the centralized portion of a Cloud Radio Access Network (C-RAN), comes

from co-locating many sets of Radio Equipment Control elements. According to the China Mobile Research

Institute, Cloud RANs will “improve network quality and coverage,

reduce transmission resource consumption and lower OPEX by

up to 50% and CAPEX by 15%”. The idea is to use consolidation

to eliminate potentially underutilized equipment. Consider the

classic city and suburb scenario. The base stations in the city are

operating near full capacity during the business day while the

sites in the suburbs are relatively lightly used because “everyone”

is at work. This flips in the evening when workers return to the

suburbs. There is no way to balance the processing load because

the processing elements in the base stations are dedicated to the

local antenna array. Consolidating the radio equipment control

has the potential to improve the balance because a single set of

Radio Equipment Controllers supports both the local (city) and the remote (suburban) Radio Equipment. In

theory even the peak load requirements should require less capacity in a cloud-based topology compared to

the classic topology.

The theoretical reduction in processing capacity presumably results in reductions in energy consumption.

However, some cloud base station concepts assume that the server facility uses classic general-purpose

servers (x86 architecture) to replicate the baseband, transport and control processing that classic base

Cloud base station positioning

Reality check

3Texas Instruments

stations perform using specialized wireless SoCs. Use of general-purpose servers is impractical on several

fronts. First, base station SoCs consume one-tenth the power of a typical server chip measured MIPS-to-

MIPS or core-to-core, but this is only part of the story. The x86 architecture’s lack of wireless accelerators

and signal processing specialization means that the functions that are traditionally performed by highly

efficient wireless SoCs need to be replicated in newly created “soft accelerators” and then road tested and

validated. The x86 architecture is positioned by its proponents as being “industry standard”. This may be

true for many general computing applications but it is not so for wireless base stations where x86 has no

appreciable market share or field experience. Base station processing is dominated by signal processing. The

industry standard in this market, as evidenced by market share is signal processing oriented SoC s from TI.

In contrast to the x86 architecture, KeyStone base station SoCs have dedicated processing hardware

implemented as AccelerationPacs that are equal to 250 GMIPs of processing that has been offloaded from

the cores. Wireless baseband, packet, transport and security processing are other examples of complex,

MIPS-intensive functions that have been incorporated into KeyStone AccelerationPacs. AccelerationPacs are

not only faster at signal processing and packet handling than programmable solutions; they are also orders

of magnitude more power efficient. Lastly, KeyStone SoCs are heterogeneous multicore devices. They contain

Digital Signal Processing/ Vector Signal Processing cores (DSP/VSP) and RISC cores along with the Accelera-

tionPacs so that they bring the best processing to the task at hand leading to solutions that are properly bal-

anced and optimized for performance and power consumption. Portions of layer 1 and 2 are differentiated by

base station vendor software so they are not part of the accelerator offload. These functions are typically pro-

cessed by DSP/VSP cores. Complex scheduling and MIMO antenna processing will become common place

for LTE. These are examples of functions where large matrix-based algorithms that are the “secret sauce” of


Introducing TI’s KeyStone multicore

architecture

• TeraNet• Multicore Navigator• Multicore Shared Memory

Chip Infrastructure

• ARM, DSP/VSP

CorePacs

• Packet processing• Security processing• Radio processing

AccelerationPacs

High performance I/O

Tera

Net

Tera

Net

Multicore NavigatorQueue Manager

I/O

Packet and Security Coprocessors

Bit Rate Radio Coprocessors

Symbol Rate Radio Coprocessors

DSP/VSP CorePacs ARM CorePacs

DSP/VSP CorePacs ARM CorePacs

Figure 2. Generic illustration of TI’s KeyStone SoC architecture.

4 Texas Instruments

base station developers dominate. Non-standard algorithms like these which are differentiated should remain

in the programmable domain. They are best handled by TMS320C66x DSP cores which have been enhanced

to support both fixed- and floating-point math. ARM RISC cores are used for base station control and man-

agement processing. TI’s KeyStone multicore architecture is the result of years of field experience in wireless

base stations focusing on delivering high performance at the lowest possible power. Figure 2 on the previous

page is a generic illustration of TI’s KeyStone SoC architecture.

Many wireless base station applications can be addressed with all-inclusive SoCs encompassing Layers 1,

2 and 3 along with Transport and control processing. But this kind of “base station on a chip” is still beyond

reach for larger macro and large scale cloud RAN requirements. For this reason, the KeyStone architecture

is designed to be inherently adaptable to different system partitioning solutions. The structural elements,

TeraNet, Multicore Navigator and the Multicore Shared Memory Controller, provide a common on-chip infra-

structure and are the critical elements enabling KeyStone devices to deliver full multicore entitlement. ARM®

CorePacs, DSP CorePacs, AccelerationPacs and I/O interfaces are attached to this KeyStone infrastructure.

This elegant architecture allows TI to generate devices to address many specialized requirements and system

partitions. Figure 3, below, depicts three common device configurations.

Once the system partitioning is determined and chips are created and installed on boards, the KeyStone

infrastructure works to make large scale systems with many devices function as a whole. Hyperlink or


KeyStone’s flexible partitioning and

system interconnect

DSP CorePacsL1, L2

Tera

Net

Multicore Navigator

I/O

DSP CorePacsL1, L2




ARM CorePacsTransport + Control

Tera

Net

Multicore Navigator

I/O





DSP CorePacsL1, L2

Tera

Net

Multicore Navigator

I/O





Figure 3: Three common device configurations of the KeyStone architecture.

5Texas Instruments

Ethernet (with integral wire-rate switching) provides the physical interconnect. Multicore Navigator provides

logical interconnection with the hardware-assisted software feature described below extending across mul-

tiple devices.

Historically it has been very challenging to write software that can scale across a range of devices where core

counts and hardware accelerators vary widely from device to device. KeyStone changes this with the intro-

duction of Multicore Navigator to ease the software scaling challenge. A large-scale base station designed for

a cloud RAN implementation will benefit greatly from Multicore Navigator. The hardware in the base station

will consist of a combination of devices ranging from devices that focus on radio processing to devices

providing transport and packet processing. Using a mix of KeyStone-enabled devices allows TeraNet and

Multicore Navigator to physically and logically interconnect the devices so that they appear to the software

as a single integrated SoC. Software scaling and load balancing are atomically implemented by Multicore

Navigator.

Multicore Navigator is an innovative new hardware implementation that is designed to simplify multicore

software development. It uses thousands of queues to provide the software an abstraction and isolation layer

that hides the details of the specific device. Given the complexity of the software and the varying processing

elements in a multicore SoC, this has become a necessity. The result is a unique approach that results in

hardware-assisted software.

Software that leverages Multicore Navigator’s hardware assist is written as small tasks rather than mono-

lithic functions. Multicore Navigator is designed to autonomously manage these tasks. The software task’s

resource needs are identified generally, as needing FFT or DSP functionality for example, rather than explicitly

targeting a specific hardware feature or a specific core. Descriptors provide this functional identity. The task

and data are queued and then the hardware autonomously manages the processing from there. Thus, moving

from a two-core to an eight-core SoC or even accessing processing resources on an adjacent device does

not require software changes. KeyStone’s Multicore Navigator queuing and descriptor system automatically

manages the transition. This is an innovative solution that uses high-performance hardware to address what

is traditionally a software challenge; the need to develop software that is both saleable and load balanced.

Mobile users do not migrate en-mass from one standard to another. Because of this, operators cannot throw

a switch and change their networks from 3G to 4G overnight. Operators need

a migration strategy. One option is to operate parallel equipment but obvi-

ously this isn’t cost efficient. A better approach utilizes multimode base station

equipment. A well implemented cloud RAN can ease this migration by coupling

efficient accelerators that address the required standards with a virtualization

layer that allocates resources to standards as necessary. Multicore Navigator’s


Multicore Navigator: Hardware-assisted

software

Managing the standards evolution

with virtualization

6 Texas Instruments

descriptors and queues are ideal for virtualization because they automatically direct the flow to the appropri-

ate processing element greatly simplifying the workload of the application layer. The real challenge comes

when the work of separate development teams, a WCDMA team and an LTE team for example, is merged

into a single runtime. KeyStone’s DMA and interrupt controllers can be managed by a virtualization layer

that reduces the integration and rework time required to merge the applications. The ideal scenario uses

virtualization that blends the hardware elements, Multicore Navigator, DMA and interrupt controllers, with a

software-based virtualization abstraction.

The hidden gem in cloud RAN deployments may ultimately be spectral efficiency gains as high speed. High-

capacity user equipment (UE) scheduling becomes an enhanced interference mitigation technique. Co-loca-

tion of large-scale antenna processing may also lead to very efficient coordinated interference management.

Rather than coordinate across multiple geographically disperse base stations, a concentrated view of the

available spectrum eliminates the need to transmit coordination data sets. The ability to generate and evalu-

ate multiple transmission scenarios between transmission intervals ultimately will determine the spectral ef-

ficiency for the entire geographic coverage area. A well designed scheduler implemented in a cloud RAN will

lead to better user scheduling and ultimately to enhanced spectrum usage. This presupposes a significant

increase in signal processing capability that is scaled to handle the larger spectral input and increase in the

number of users. Performance is critical here and TI’s C66x DSP core excels at the matrix calculations and

data set sorting that is required.

There are a few potential obstacles that may prevent the vision of cloud RAN from becoming a widespread

reality. Probably the largest is the cost and availability of long-distance fiber. Often wireless or other wired

links are used for base station backhaul. This is because fiber may be impractical for cost, physical impedi-

ments or right-of-way issues. These alternatives to fiber simply do not have the bandwidth to transport the

volume of antenna data that a cloud base station requires. Where these impediments exist, traditional macro

or small cell base stations will continue to be deployed. What percentage of the overall footprint this turns out

to be remains to be seen.

Consideration must be given to the absolute power consumed by a cloud base station, comparing the

reduction in the field with the new equipment at the server site before cloud base stations get the green

energy seal of approval. Replacing relatively efficient base stations in the field with inefficient servers, relying

on “soft acceleration” is likely to result in a net increase in overall power consumption. This was discussed

earlier but it bears repeating because power savings is often a key attribute cited to justify moving to cloud

base stations. “Soft acceleration” as may be performed on x86 servers cannot match either the performance

or the power efficiency of KeyStone’s hardened AccelerationPacs. KeyStone’s balance of programmable and

Scheduling in the cloud

Clouds on the horizon?


7Texas Instruments

accelerated processing elements provides the ideal balance between performance and power consumption

for both traditional base stations and Cloud RAN.

The final potential obstacle that requires further study is latency. This is particularly true with LTE where

latency budgets are stringent. The philosophy of LTE is in part to push processing out to the field in order

to improve performance over 2G and 3G systems. Propagation delay over relatively long fiber links may not

be high, but it might induce enough latency to negatively impact spectral efficiency. This alone could hinder

cloud RAN deployments or limit them to dense city environments where the distance to the server facility will

be short.

TI has been on an accelerated integration path creating KeyStone SoCs with this goal in mind. KeyStone

provides the required processing power and efficient wireless baseband acceleration needed to achieve the

“green” power footprint and operational benefits of a Cloud RAN. Multicore Navigator offers the programming

unification and functional partition flexibility that large scale, centralized base stations will require. Now that

these integrated SoCs exist, our sights are set on enabling large scale cloud base stations which actually

require disintegration. The optimal partitioning of processing for a cloud base station may not be every

function in clusters of identical SoCs. We are currently exploring with our customers the optimum system

partitioning for cloud base stations. Figure 4 depicts one example where KeyStone transport and control

devices are clustered with KeyStone baseband devices with Multicore Navigator providing logical and physical

interconnection.

The critical point is that regardless of what the optimal or manufacturer-preferred partitioning of the sys-

tem turns out to be, the physical and logical extensibility of TI’s KeyStone architecture will apply. The resulting

Conclusion




I/O I/O I/O

I/O I/O I/O

Tera

Net

TeraNet via HyperLink or

Ethernet provides physical

connection

Multicore Navigator

provides the logical

connection for scalable software

DSP CorePacsL1, L2

QueueManager

Tera

Net

Multicore Navigator

DSP CorePacsL1, L2




DSP CorePacsL1, L2

QueueManager

Tera

Net

Multicore Navigator

DSP CorePacsL1, L2




DSP CorePacsL1, L2

QueueManager

Tera

Net

Multicore Navigator

DSP CorePacsL1, L2




DSP CorePacsL1, L2

QueueManager

Tera

Net

Multicore Navigator

DSP CorePacsL1, L2




DSP CorePacsL1, L2

QueueManager

Tera

Net

Multicore Navigator

DSP CorePacsL1, L2




ARM CorePacsL3 + Control

Tera

Net

Multicore Navigator

ARM CorePacsL3+ Control


Tera

Net

Multicore Navigator



Tera

Net

Multicore Navigator

Tera

Net

Tera

Net

Tera

Net

I/O








Multicore Navigator

Multicore Navigator

Multicore Navigator

I/O I/O

DSP CorePacsL1, L2

DSP CorePacsL1, L2

Figure 4: Example of KeyStone transport and control devices clustered with KeyStone baseband devices with

Multicore Navigator providing logical and physical interconnection.


6 Texas Instruments

SPRY183© 2011 Texas Instruments Incorporated

Important Notice: The products and services of Texas Instruments Incorporated and its subsidiaries described herein are sold subject to TI’s standard terms and conditions of sale. Customers are advised to obtain the most current and complete information about TI products and services before placing orders. TI assumes no liability for applications assistance, customer’s applications or product designs, software performance, or infringement of patents. The publication of information regarding any other company’s products or services does not constitute TI’s approval, warranty or endorsement thereof.

C6000, C64x+ and DaVinci are trademarks of Texas Instruments Incorporated. All other trademarks are the property of their respective owners.

wireless processing for cloud RAN or super macro cell will perform as an integrated unit with hardware

assisted, scalable software. TI’s KeyStone architecture is an excellent choice for the construction of cloud

base stations.

IMPORTANT NOTICE

Texas Instruments Incorporated and its subsidiaries (TI) reserve the right to make corrections, modifications, enhancements, improvements,and other changes to its products and services at any time and to discontinue any product or service without notice. Customers shouldobtain the latest relevant information before placing orders and should verify that such information is current and complete. All products aresold subject to TI’s terms and conditions of sale supplied at the time of order acknowledgment.

TI warrants performance of its hardware products to the specifications applicable at the time of sale in accordance with TI’s standardwarranty. Testing and other quality control techniques are used to the extent TI deems necessary to support this warranty. Except wheremandated by government requirements, testing of all parameters of each product is not necessarily performed.

TI assumes no liability for applications assistance or customer product design. Customers are responsible for their products andapplications using TI components. To minimize the risks associated with customer products and applications, customers should provideadequate design and operating safeguards.

TI does not warrant or represent that any license, either express or implied, is granted under any TI patent right, copyright, mask work right,or other TI intellectual property right relating to any combination, machine, or process in which TI products or services are used. Informationpublished by TI regarding third-party products or services does not constitute a license from TI to use such products or services or awarranty or endorsement thereof. Use of such information may require a license from a third party under the patents or other intellectualproperty of the third party, or a license from TI under the patents or other intellectual property of TI.

Reproduction of TI information in TI data books or data sheets is permissible only if reproduction is without alteration and is accompaniedby all associated warranties, conditions, limitations, and notices. Reproduction of this information with alteration is an unfair and deceptivebusiness practice. TI is not responsible or liable for such altered documentation. Information of third parties may be subject to additionalrestrictions.

Resale of TI products or services with statements different from or beyond the parameters stated by TI for that product or service voids allexpress and any implied warranties for the associated TI product or service and is an unfair and deceptive business practice. TI is notresponsible or liable for any such statements.

TI products are not authorized for use in safety-critical applications (such as life support) where a failure of the TI product would reasonablybe expected to cause severe personal injury or death, unless officers of the parties have executed an agreement specifically governingsuch use. Buyers represent that they have all necessary expertise in the safety and regulatory ramifications of their applications, andacknowledge and agree that they are solely responsible for all legal, regulatory and safety-related requirements concerning their productsand any use of TI products in such safety-critical applications, notwithstanding any applications-related information or support that may beprovided by TI. Further, Buyers must fully indemnify TI and its representatives against any damages arising out of the use of TI products insuch safety-critical applications.

TI products are neither designed nor intended for use in military/aerospace applications or environments unless the TI products arespecifically designated by TI as military-grade or "enhanced plastic." Only products designated by TI as military-grade meet militaryspecifications. Buyers acknowledge and agree that any such use of TI products which TI has not designated as military-grade is solely atthe Buyer's risk, and that they are solely responsible for compliance with all legal and regulatory requirements in connection with such use.

TI products are neither designed nor intended for use in automotive applications or environments unless the specific TI products aredesignated by TI as compliant with ISO/TS 16949 requirements. Buyers acknowledge and agree that, if they use any non-designatedproducts in automotive applications, TI will not be responsible for any failure to meet such requirements.

Following are URLs where you can obtain information on other Texas Instruments products and application solutions:

Products Applications

Audio www.ti.com/audio Communications and Telecom www.ti.com/communications

Amplifiers amplifier.ti.com Computers and Peripherals www.ti.com/computers

Data Converters dataconverter.ti.com Consumer Electronics www.ti.com/consumer-apps

DLP® Products www.dlp.com Energy and Lighting www.ti.com/energy

DSP dsp.ti.com Industrial www.ti.com/industrial

Clocks and Timers www.ti.com/clocks Medical www.ti.com/medical

Interface interface.ti.com Security www.ti.com/security

Logic logic.ti.com Space, Avionics and Defense www.ti.com/space-avionics-defense

Power Mgmt power.ti.com Transportation and Automotive www.ti.com/automotive

Microcontrollers microcontroller.ti.com Video and Imaging www.ti.com/video

RFID www.ti-rfid.com

OMAP Mobile Processors www.ti.com/omap

Wireless Connectivity www.ti.com/wirelessconnectivity

TI E2E Community Home Page e2e.ti.com

Mailing Address: Texas Instruments, Post Office Box 655303, Dallas, Texas 75265Copyright © 2011, Texas Instruments Incorporated

http://www.ti.com/audio

http://www.ti.com/communications

http://amplifier.ti.com

http://www.ti.com/computers

http://dataconverter.ti.com

http://www.ti.com/consumer-apps

http://www.dlp.com

http://www.ti.com/energy

http://dsp.ti.com

http://www.ti.com/industrial

http://www.ti.com/clocks

http://www.ti.com/medical

http://interface.ti.com

http://www.ti.com/security

http://logic.ti.com

http://www.ti.com/space-avionics-defense

http://power.ti.com

http://www.ti.com/automotive

http://microcontroller.ti.com

http://www.ti.com/video

http://www.ti-rfid.com

http://www.ti.com/omap

http://www.ti.com/wirelessconnectivity

http://e2e.ti.com

creating cloud base stations with ti s keystone … called the cloud radio access network (c-ran),...

Documents