Making Sense of O-RAN and vRAN - Part Two
In my previous blog I explained the basic definitions of Open RAN and Virtualized RAN (vRAN), while also describing the industry’s positive expectations and the challenges ahead. Now, I want to tell you about why Nokia believes in Open RAN and why silicon technology continues to be a key enabler for RAN solutions, even in an open ecosystem with more open interfaces.
First, let’s address a question I am often asked. Why is Nokia committed to Open RAN? Well, being an established and leading radio networks supplier, we enjoy the benefits of a large installed base and a long list of customers. The market is very competitive, and it could be tempting for a supplier like us to exploit a proprietary RAN ecosystem for the maximum time possible. However, Nokia decided to endorse the foundation of the original Open RAN consortium when it consisted of only five operators. We have since contributed to many Open RAN working groups. One of our key contributions was the eCPRI7-2 specification. In July 2020, we shifted gears, shared more ambitious Open RAN roadmaps and release plans, and joined the Open RAN Policy Coalition. Some people have suggested we are shooting ourselves in the foot with this, but I guarantee this is not the case. In fact, these moves make complete sense for Nokia. Here’s why.
First of all, make no mistake, Open RAN will happen. It will happen with or without us, and it will happen with or without some of our traditional competitors. The question is whether we should just wait and see, or actively contribute to Open RAN and shape our own destiny. We opted for the latter. This should be no surprise to our customers. They say, almost without exception, that “you guys have great ideas, you innovate and try to create something new, whereas the other guys are conservative and try to protect the past.”
Why will Open RAN happen?
Now, why do we think Open RAN will happen? Here are some reasons:
- Customers want it – Major operators are pushing for open solutions to reap the benefits as soon as possible. Last time I checked, there were 27 operators in the Open RAN Alliance. As many as 23 of these are among our largest RAN customers. They are true leaders with strong technical competence.
- Governments want it – Some countries (and operators) are looking at Open RAN as a way to introduce new suppliers to the market, as Open RAN dramatically reduces entry barriers for new players. For some countries, it is important to have more options if some suppliers have to be excluded or limited in their presence. Some countries are even looking at Open RAN as a mechanism to drive local 5G innovation. We take no part in politics, but these are the geopolitical seas upon which we must sail.
- It is an opportunity to drive innovation – An Open RAN ecosystem will accelerate innovation due to a wider diversity of players and to the nature of the open architecture. RAN Intelligent Controller (RIC) - a new network function exposing an open API for AI/ML - is a great example of this. 5G is creating new possibilities in both evolved mobile broadband and ever better Internet-of-Things (IoT) with URLLC and eMTC functions. Open RAN will further catalyze this process.
- Cloud computing enables Open RAN and reduces entry barriers – New entrants making Open RAN baseband cannot afford to develop custom silicon, nor should they. Cloud computing continues to evolve to be more feasible and competitive for vRAN.
With all the above and more considerations, it is clear there are enough reasons for the market to evolve to open standards and that Nokia’s best move is to lead the transformation. We believe the Open RAN ecosystem needs a strong player to help drive this process and Nokia, with its long history in supporting open initiatives, is working to take this leadership position.
How will Open RAN come about?
Let’s now dive a little bit more into the “How”. While Open RAN provides flexibility in supplier selections, it brings some challenges in end-to-end radio optimization. Nokia studies show that today the best radio performance can be achieved by using a Massive MIMO (M-MIMO) active antenna and baseband from the same supplier because it enables more processing to be integrated into the M-MIMO antenna. It is conceivable that the ultimate radio coverage and capacity performance are achieved with full, end-to-end Nokia RAN solution and with optimized fronthaul interface between radio and baseband. However, we recognize that in an open environment, multi-vendor scenarios will be common and thus we must serve and optimize them.
In RAN disaggregation cases where Nokia is selected as the preferred baseband or radio unit supplier, we will ensure that operators get high quality radio performance with Nokia-recommended interoperability profiles. These profiles represent the Nokia Open RAN product strategy to support open interfaces and integration with third party products. A comprehensive RAN feature set makes the end-to-end Nokia O-RAN solution complete. Many of our customers may want to use all-Nokia products in dense urban areas (Open RAN compliant, of course!) and a mix of Open RAN suppliers in other coverage areas. Such a mix may or may not include Nokia Radio Units (RU), Distributed Units (DU) or Centralized Units (CU).
Another point I mentioned in my previous blog was about the implementation decisions needed to optimize the development (and performance) of Open RAN. One point of discussion has been a perceived contradiction between building Open RAN solutions and investing in System on Chip (SoC) development. Let’s split this in two parts: radio (RU) and baseband (DU, CU).
For the foreseeable future, we will need custom SoCs for RU Digital Front Ends (DFE). You can’t really virtualize the DFE. In M-MIMO configurations, we will also need custom SoCs for so-called L1-Low or digital beamforming. It therefore makes sense for Nokia to continue investing in its ReefShark SoC family for DFE and L1-Low (for M-MIMO).
Let’s move to the baseband. L2 (especially non-real-time sensitive L2) and L3 processing can be done reasonably well with General Purpose Processors (GPPs) in any Commercial off the Shelf (COTS) cloud computing platform. This means the CU is the first candidate to be virtualized, and this is what we call vRAN1.0 (vCU + DU).
It must be said that today’s GPP and associated FPGA- or eASIC-based hardware acceleration solutions are not particularly efficient in the heavy-lifting of L1 processing in the DU. Custom SoCs outperform GPPs approximately tenfold in CAPEX and power consumption per unit of cell throughput, cell connectivity and subscriber connectivity. But be careful! A recent business case study we completed with a major carrier in a large market with good fiber infrastructure revealed that the overall benefits of cloud computing in the vRAN use case outweighed the cost disadvantages of cloud computing in vRAN L1 CAPEX efficiency and power consumption. It is important to always look at the bigger picture!
Consequently, there is now a race in the market to the best vRAN L1 SoC. We expect different suppliers will try to become the choice of vRAN L1 SoCs on hardware acceleration boards that operators and webscalers (whose public service clouds operators may want to use) will put into their data centers.
How will vRAN evolve?
Now, let’s look at the vRAN evolution roadmap.
Our first vRAN1.0 solution consisted of “bare metal” DU with custom SoC and virtualized CU (vCU) based on currently available GPPs.
Nokia’s vRAN2.0 solution implements a virtualized DU based on GPPs complemented by eASIC look-aside hardware accelerators. These accelerators are used primarily for Forward Error Correction (FEC).
In parallel with vRAN2.0, Nokia Mobile Networks and Bell Labs have been researching other processing environments, especially for L1. One alternative we often refer to as “vRAN3.0” is to use Graphics Processing Units (GPUs) for the L1 portion of processing in the vDU. While a GPU-based solution could potentially offer higher CAPEX efficiency and power efficiency than GPP with hardware accelerators, it is still not as efficient as custom SoCs in so-called “bare metal” solutions. The GPU-based solution does, though, offer flexibility with respect to network slicing and support of other GPU-based flows for application providers including AI/ML workflows, image and object recognition as well as other enterprise applications.
As we don’t think vRAN3.0 will necessarily be the “end-game”, we are researching what we call “vRAN4.0”.
A truly open solution
Some new Open RAN or vRAN entrants have been flinging mud at Nokia by claiming that our vRAN solution would not be open. Nothing could be further from the truth. We are the most open, flexible and versatile Open RAN and vRAN supplier.
Let’s start with the vertical dimension. An operator can buy the full stack from us - Nokia AirFrame Open Edge cloud computing hardware platform; Nokia cloud infrastructure software stack; and Nokia vRAN software. But an operator can also buy its favorite COTS cloud computing hardware from someone else; buy their favorite cloud infra software stack from someone else; and run Nokia vRAN software on it.
Our AirFrame Open Edge servers for vCU and vDU are COTS and based on GPPs. Whatever solution we use in vRAN L1 going forward, it will have to be something that operators and webscalers accept as an open and standard solution.
In the horizontal dimension, we are of course completely open to supply any combination of RU, RAP (RU+DU), DU, and CU. Let me repeat: you don’t have to buy them all from Nokia, but you are of course welcome to do so. The first priority of our customers is to have the front-haul interface open, so that two different suppliers’ RU and DU/CU can be combined. But later, the DU and CU can be from different suppliers, and we will even split CU into CU-UP (User Plane) and CU-CP (Control Plane). May the best win!
Summing up, I have covered why Open RAN will happen, why Nokia is investing in Open RAN, and what steps Nokia is taking to assume a leadership position in this open ecosystem. In addition, I have explored the significant role of processing solutions in the future Open RAN ecosystem, and why it still makes a lot of sense for Nokia to continue investing in its ReefShark SoC family.
The story doesn’t end here. As Open RAN is such a hot topic, this series will continue to address all the most relevants topics related to our industry’s journey to openness. Stay tuned for the next blogs.