Skip to content
5G/6G AcademyEst. 2009

O-RAN vs vRAN

People use the two words as if they meant the same thing. They don't, and the difference matters the moment you're asked to plan, buy or troubleshoot one.

vRAN (virtualized RAN) is about where the software runs. The baseband — the DU and CU work that used to live in a vendor's dedicated box — runs as software on ordinary servers, usually with an accelerator card for the heaviest Layer 1 processing. O-RAN (Open RAN) is about who can talk to whom. It defines open interfaces between the radio unit, the DU, the CU and a RAN Intelligent Controller, so a radio from one vendor can work with a DU from another.

You can have one without the other. A single vendor can sell you a fully virtualized RAN with closed interfaces, and an O-RAN radio and DU can run on purpose-built hardware. In practice most new Open RAN networks are both — open interfaces and software on cloud infrastructure — which is why the names get blurred.

AspectvRANO-RAN
What it changesWhere the baseband software runs: general-purpose servers instead of dedicated hardwareThe interfaces between RAN parts: opened and specified so vendors can be mixed
Who defines itNot a standard in itself; built on 3GPP's CU/DU split and a cloud platformThe O-RAN ALLIANCE, on top of 3GPP (Open Fronthaul, E2, A1, O1, O2)
More than one vendor?Not required — often one vendor's software on commercial serversThe whole point: radio, DU, CU and RIC can each come from a different vendor
HardwareCommercial servers, usually with an L1 accelerator (look-aside or inline)Anything; the radio unit is always specialized hardware
FronthaulWhatever the vendor uses, which can be proprietaryOpen Fronthaul: split 7.2x over eCPRI, with defined control, user, sync and management planes
RAN intelligenceNot part of the ideaNear-RT RIC (xApps, 10 ms–1 s loops) and non-RT RIC (rApps, slower than 1 s)
Main winOff-the-shelf hardware, pooling, faster software upgradesVendor choice, price competition, a programmable RAN
Main headacheMatching dedicated silicon on Layer 1 speed and power drawIntegration: when it breaks, which vendor owns the fault?

vRAN: the same RAN, running as software

A classic site has a baseband unit with the vendor's own chips doing everything from channel coding to scheduling. vRAN takes the functions 3GPP already split into a CU and a DU and runs them as software — these days usually as containers on a Kubernetes platform.

The CU is the easy part. It handles RRC and PDCP, nothing in it has to happen inside a slot, and it virtualizes happily. The DU is harder: its MAC scheduler and high-PHY work to sub-millisecond deadlines, and Layer 1 jobs like LDPC decoding eat CPU. That's why almost every vRAN DU has an accelerator. In a look-aside design the CPU hands only the forward error correction to the card; in an inline design the whole Layer 1 pipeline runs on it.

O-RAN: opening the boxes

3GPP had already split the gNB into a CU and a DU joined by the standard F1 interface. The O-RAN ALLIANCE, formed in 2018, went further: an open fronthaul between the radio unit (O-RU) and the DU (O-DU) using split 7.2x, plus the management and control interfaces needed to run a mixed-vendor network — O1 for management, O2 for the cloud platform, and E2 and A1 for the RAN Intelligent Controllers.

The RIC is O-RAN's own addition. The near-real-time RIC runs small apps (xApps) that adjust things like handover or traffic steering on 10 ms to 1 s loops. The non-real-time RIC sits in the management layer and runs rApps for slower jobs: policy, analytics, training ML models.

Where they meet, and what it means on the job

Most Open RAN builds are virtualized too. Rakuten Mobile in Japan was the first large network built that way, launching commercially in 2020. Plenty of operators also take vRAN from a single vendor first and open the interfaces later.

The day-to-day work changes either way. On a vRAN DU you end up caring about things a box hid from you: CPU pinning, real-time kernels, NUMA placement and accelerator drivers. On a multi-vendor O-RAN site the usual suspects are on the fronthaul: PTP timing on the sync plane, mismatched M-plane configuration, or the two ends disagreeing on the compression bit-width. Knowing which layer to look at first is most of the job.

The bottom line

The two answer different questions. vRAN asks can we stop buying special boxes? O-RAN asks can we stop buying everything from one vendor? Operators who want both end up with what the industry calls Open vRAN.

If you're building a career on the radio side, learn them together: every O-RAN deployment you'll touch is very likely virtualized, and every vRAN team is being asked whether it can open up.

Frequently asked questions

Is O-RAN the same as vRAN?
No. vRAN is about running baseband software on general-purpose servers; O-RAN is about open interfaces between radio, DU, CU and RIC so different vendors can be combined. Many networks do both.
Can you have vRAN without O-RAN?
Yes. A single vendor can supply virtualized CU and DU software on commercial servers while keeping its own interfaces between radio and baseband.
What is split 7.2x?
The O-RAN fronthaul split. The radio unit keeps the low-PHY (FFT and, in most designs, beamforming), and the DU does the high-PHY, MAC and RLC. It keeps fronthaul bandwidth manageable without making the radio unit too complex.
Is C-RAN the same as vRAN?
No. C-RAN (centralized RAN) is about pooling basebands in one hub site; vRAN is about running them as software. They're often combined, because software makes pooling easier.

Get one 5G/LTE engineering deep-dive a fortnight

One thing explained properly every other Tuesday — a real failure, and the lab or trace where you work it yourself. No spam, no marketing, just engineering. Unsubscribe in one click.