Open radio access network (ORAN) is a concept that has been in development for a while, aiming toward open interfaces and making the radio access network more flexible and vendor-neutral. However, despite a few successful greenfield trials, it appears ORAN is not yet ready for full-fledged commercial deployment. Several factors, such as technology maturity, security, and energy efficiency, are new add-ons to existing reasons. Traditional OEM players in the RAN market restrict the interoperability of their inter-twining components, forcing operators into vendor lock-in. Operators must depend on E2E equipment from one exclusive vendor. Thus, without open interoperability standards being enforced, entry for new vendors at any level of the RAN market is challenging. Open RAN offers a more flexible, cost-effective, and innovative approach to building and operating 5G networks.
Challenges in ORAN Hardware and Deployment
The gNodeB is split into a Central Unit (CU), Distributed Unit (DU), and Radio Unit (RU) called O-CU, O-DU, and O-RU in ORAN specifications. The CU is further split into two logical components, one for the Control Plane (CP) and one for the User Plane (UP). This logical split allows different functionalities to be deployed at different locations of the network and on different hardware platforms. Managing the technical complexity and rapid technology innovation involved in a multi-vendor component environment and seeking T-shirt-size tailor-made deployments in ORAN are critical factors. T-shirt sizing is an approach toward custom design-sizing of ORAN software and network hardware infrastructure based on a particular user base or network capacity. Integrating various scenarios using these components for an effective pre-integrated rollout for deployment seems a distant dream for an operator. A single-vendor, integrated infrastructure is more practical today, given the ease of deployment. Most commercial ORAN deployments are hardware-dependent, with a few chosen OEM vendor products coupled with hardware accelerators to manage the performance shortfalls of x86 General Purpose Processor (GPP) hardware.
ORAN aims to accelerate the RF PHY layer with a GPP hardware card, but the current GPP hardware cannot fulfill the performance glitches at baseband processing. Here, we cannot plug a specialized AMD or ARM chipset into the GPP hardware to manage baseband data processing and performance. ORAN hardware architecture also demands network and software separation to align with the open architecture standards, but this baseband processing cannot be cross ported between GPP, SPP hardware and vice-versa. These limitations make it challenging for operators to deploy ORAN to support the separation of hardware and software.
On the other hand, the 5G window of opportunity is slowly closing, and telcos in Europe are moving toward the deployment of 5G Advanced or 6G as a concept, given their challenging financial situation. Although forums such as 3GPP, TIP, Open Air Interface, ORAN Alliance, and Open RAN Policy Coalition are working to fix ORAN integration issues, progress has been slow. Most telcos that have deployed ORAN are still stuck with traditional OEM vendor solutions, quite far from the idea of separation of hardware and software with a high degree of interoperability.
Streamlining Technology Processes in ORAN System Integration
ORAN specifications aim to make system integration simpler with plug-and-play-like deployment components. However, operators must either integrate themselves or look for a smart system integrator to T-shirt-size RAN and engineer or integrate it. There is a dearth of skilled resources who can custom design an ORAN implementation with interoperability from a design skills perspective. An open RAN system is all about having a disaggregated, software-defined, virtualized functionality built using open interface specifications implemented in vendor-neutral GPP hardware. However, 5G ORAN has a long way to go before it can be fully deployed commercially. Adopting higher bandwidth and MIMO (massive input, massive output) requires a more efficient and enhanced CPRI (common public radio interface) interface with more functional splits to reduce the overall front haul (FH) bandwidth requirements.
One of the challenges of evolved common public radio interface (eCPRI) specifications is that it is not defined as an open and interoperable interface. Hence this calls for the DU (distributed unit) and RU (radio unit) to be procured from the same supplier for seamless integration. However, from an ORAN design perspective, an open front haul interface that enables the ORAN distributed unit (o-DU) and ORAN radio unit (o-RU) should ideally be sourced from different vendors. An essential pre-requisite for open fronthaul-based solution adoption and commercial deployment is to bring performance and capabilities on par with integrated solutions. Hence a disaggregated solution must support wide bandwidth, multi-band and high output power requirements, low latency, low power consumption, and functional requirements such as network sharing, dynamic spectrum sharing, etc.
Managing an Effective ORAN HW Rollout
Alignment between all RAN components is required to ensure interoperability with all Open RAN technical requirements. A consistent approach across all RAN components requires strong adherence to standards and specifications. Slight variations in terms of inconsistencies may impact RAN stability and performance. Although ORAN specifications aim to simplify system integration, they eventually drive toward attempting to deploy ORAN in a plug-and-play environment. Telco operators must integrate themselves or look for a skilled system integrator to implement ORAN with a vendor-neutral hardware-based approach. The disaggregation of a hardware base station implies reaggregation with the necessary software, hardware, and separation of the plane ecosystem with virtualization based on an optimized cost model telcos can afford. However, with ORAN a long way from commercial deployment, telcos may need to continue relying on traditional vendor solutions or adopt a cautious approach toward ORAN deployment. The O-RAN Alliance has developed interoperable interfaces between various parts of the network. However, these standards are too open; hence, we will require a mix of system engineering and integrators to design, develop, and integrate Open RAN subcomponents. In Massive Input Massive Output (MIMO), Rx (receiver) and Tx (transmitter) antennas are stuffed into the frame and placed with more transmitters and receivers. Performance is seen to improve when hardware and software are tightly coupled.
Technology Comparison: GPP vs. FPGA vs. ASIC
x86-GPPs are designed with a high clock rate to implement a large instruction set of operations sequentially. Generally, they have about 6 to 15 times higher clock rate than FPGA (field programmable gate array) implementations and about 6 to 8 times higher clock rate than a GPU. x86-GPPs are better known to handle and solve general-purpose problems. Each architecture is most efficient for a problem that matches its instruction and data granularity. For a problem with 32-bit data that maps well to the instructions available in a given Industry Standard Architecture (ISA) and that does not have a lot more inherent parallelism than what the available ISA implementation offers, a general-purpose CPU is excellent.
By comparison, an ASIC (application-specific integrated circuit) implementation can exactly match the data types, instructions, and spatial parallelism that suits a problem, has a significant advantage, but is inefficient as it must make a lot of worst-case assumptions, e.g., it always implements the same number of processing elements no matter the size of the problem instance. The FPGA inherits all advantages of the ASIC implementation at a price of a factor of 10 to 22 times. For certain specific applications, the FPGA can more than compensate for that by its ability to tailor the hardware to the problem instance.
Toward a Hybrid ASIC-Powered Solution
The current use of cost-effective x86-based GPP hardware in open RAN deployments is partly responsible for the open architecture RAN system's inefficiency and performance-related issues. Based on this conclusion, we propose using hybrid HW support consisting of ASICs and FPGA as hardware accelerators to support GPP hardware to manage more efficient baseband processing between RU and DU units. An ASIC chipset HW shows better performance at baseband processing and is cheaper than using a GPP HW or chip accelerator. This means ASICs can run faster and more efficiently than GPP HW or FPGAs. ASICs can be tailored to baseband hardware processing tasks and hold only the resources required for that ORAN baseband design, with no wastage.
In conclusion, RUs are generally implemented on FPGAs and ASIC boards and deployed close to RF antennas. Owing to their highly generalized nature, FPGA accelerators are traditionally power-hungry devices. In contrast, an ASIC accelerator is built on a modern technology platform that draws such little power that it can possibly drive a project using solely harvested energy to manage CU and DU white box hardware. Intelligent hardware features can be used to boost radio performance, double throughput on its network, and reduce overall costs.
About the author
Joe is the Head of Network Engineering & Solution Development at Cyient, Technology Office with focus on Telecom Networks. He is also a Senior Member at IEEE, Member ACM and works for various forums towards standardization & compliance of Telecom Technologies. He works on engineering solutions with strong cross domain expertise and helps clients realize business value with providing n expertise in designing & developing systems to improve network experience, enabling competitive edge to our customers and thus ensure high productivity & customer profitability. He has an extensive experience in leading Carrier and Enterprise networks, Telecom Network Engineering - Disaggregation of networks, Softwarization, Virtualization and Containerization of wireless and optical networks. Managing Radio Technologies, ORAN engineering & evolution, Cross domain Network Orchestrators, Radio intelligent Controllers with AI and Data Science algorithms, OSS & BSS systems integration and engineering and implementing digital readiness on sustainable network business models.