Skip to content
Lunera Pitch Lunera

4 min read ·

What Software Defined Means—and What Founders Should Prove

Understand software-defined networking, storage and radio, then test whether programmability creates customer value and a durable startup advantage.

Share X in f
Lunera · 4 min read

Software defined describes systems in which important behavior or services are implemented or controlled through changeable software, rather than being tightly fixed to a particular hardware implementation. Hardware still does the physical work; software creates more flexibility over how that work happens.

The term is incomplete without a noun. Software-defined networking, storage and radio share a preference for programmability, but they move different functions into software. For a technical founder, the useful question is not whether a product deserves the label. It is what customers can now change, what that change costs, and why your implementation is meaningfully better.

What changes in a software-defined system?

A common architectural pattern is to separate decisions or services from the resources that execute them:

  • A software layer expresses behavior, policy or configuration.
  • An interface connects that layer to underlying resources.
  • Those resources execute the work within their physical limits.

Networking makes this separation especially clear. The control plane decides how packets should be forwarded; the forwarding plane, often called the data plane, handles packets. Software-defined networking introduces an abstraction between them so applications can program network behavior. The IRTF’s informational RFC 7426 describes this separation and the interfaces between the layers.

That model is useful, but it is not a universal specification for everything called software defined. In radio, for example, software can implement signal-processing functions themselves—not just manage a separate execution layer.

Three examples—and their limits

Domain What becomes software defined What does not disappear
Networking (SDN) Applications can program network behavior through a separation of control and forwarding functions. Packet-forwarding resources and the need to design for controller connectivity and failures. See RFC 7426.
Storage (SDS) Storage services are separated from a proprietary hardware bundle, allowing software to manage how and where data is stored. Physical disks and servers. Red Hat’s explanation explicitly distinguishes abstracting storage services from removing the hardware that stores data.
Radio (SDR) Some or all physical-layer functions run as modifiable software or firmware on programmable processors. The radio hardware needed to transmit or receive signals. The Wireless Innovation Forum includes FPGAs, DSPs and general-purpose processors among the implementation options.

The benefit is flexibility within a supported envelope, not unlimited capability. A software update is not a substitute for sufficient physical resources.

Ceph provides a concrete storage example of that limit: although designed for commodity hardware, its documentation calls for balancing failure domains, cost and performance. It also recommends resource headroom because startup, recovery and rebalancing can demand more than steady-state operation. Ceph’s hardware guidance for its Squid release makes clear why “software defined” does not mean “hardware independent” in a performance sense.

Not the same as automation, virtualization or SaaS

These distinctions are useful when describing a product:

  • Automation concerns how an operation gets performed. A script can automate existing device configuration without changing the underlying architecture.
  • Virtualization abstracts resources. It can support a software-defined architecture, but is not equivalent to one. For example, Red Hat distinguishes pooling storage capacity through virtualization from separating storage services from devices through SDS.
  • SaaS describes how software is delivered. It does not establish whether the product separates hardware-bound functions or exposes a programmable resource layer.

An API or dashboard alone is therefore weak evidence for a software-defined claim. Explain which function it changes and which coupling it removes. For broader context on where such a product sits, see the layers of software infrastructure.

What founders should prove

Treat “software defined” as an architectural claim, not a business advantage by itself. A useful evaluation has four parts.

1. Name the old constraint. Was the customer replacing appliances, configuring devices individually, or waiting for a vendor-specific capability? Identify the workflow and buyer rather than claiming general flexibility.

2. Demonstrate the new operation. Show a specific change through software, including validation and rollback. Record elapsed time, operator effort and failure behavior against the customer’s actual alternative.

3. Account for the shifted burden. Ask who now owns integration, upgrades, capacity planning and incident response. Where a separate control service exists, test what happens when it is unreachable. Test what happens when a change reaches only part of the system. Easier policy changes are not sufficient if recovery becomes harder.

4. Separate architecture from defensibility. Ask why a customer would choose your implementation over an incumbent, an open-source project or an internal tool. Candidates to test include better workload performance, broader compatibility, safer changes and lower operational effort. None becomes a durable advantage merely because it is implemented in software.

For a hypothetical storage startup, a stronger pitch than “we make storage software defined” would be: “We let this customer provision storage across its supported servers through one policy interface. Our pilot will compare provisioning effort and recovery performance with its current appliance workflow.” That statement exposes both the mechanism and the evidence still needed.

Carry that specificity into the business pitch: customer constraint, programmable operation, measured result and remaining dependency. The label tells readers where to look; the evidence tells them whether a company is worth building.