This article is based on a theCUBE interview originally published by SiliconANGLE, conducted at VMworld in San Francisco. Interviewers were Stu Miniman (@stu) and Justin Warren (@jpwarren). The interviewee is Bob Ghaffari, General Manager of the Enterprise & Cloud Networking Division at Intel Corp.
The Hardware Underneath the Cloud
“There is no cloud; it’s just someone else’s computer” is a well-worn phrase in infrastructure circles, but it carries real weight at events like VMworld. Beneath every abstraction layer — every virtual machine, container, and software-defined network — sits physical hardware. At VMworld in San Francisco, networking and storage vendors gathered to remind the industry that hardware choices still shape what is possible in multicloud architectures.
Bob Ghaffari, General Manager of Intel’s Enterprise & Cloud Networking Division, used the event to outline how Intel is positioning its networking portfolio for hybrid and multicloud environments — from foundational network interface cards (NICs) to programmable FPGA-based accelerators and edge computing solutions.
A Decade of Network Transformation: Intel’s Role in DPDK
Ghaffari traced Intel’s networking involvement back roughly ten years, to the industry’s shift away from proprietary network processors toward standard, open architectures. A pivotal contribution was Intel’s open-sourcing of the Data Plane Development Kit (DPDK) — a set of libraries and drivers for fast packet processing that runs in user space and bypasses the kernel networking stack for dramatically higher throughput.
DPDK became foundational infrastructure for software-defined networking (SDN) and network functions virtualization (NFV). By enabling high-speed packet processing on commodity x86 hardware, it helped precipitate the broader shift toward software-defined, cloud-native networking that VMware NSX and similar platforms depend on today.
“We open-sourced [DPDK], which is really this fast packet processing library that we basically enabled the industry on. And with that, there’s basically been this revolution in terms of how networking has come together.”
— Bob Ghaffari, Intel
Intel and VMware NSX: Enabling Hybrid and Multicloud Networking
Ghaffari highlighted VMware NSX as a key integration point, noting that NSX had matured significantly in the preceding year and was enabling newer, more sophisticated networking use cases. For Intel, NSX represents a direct enabler of hybrid cloud and multicloud connectivity — the ability to extend consistent networking policies across on-premises infrastructure and multiple public cloud environments.
Intel’s strategy is to offer a tiered hardware portfolio — from cost-optimized NICs to high-performance FPGA-based SmartNICs — so that operators can match hardware capability to workload requirements regardless of which cloud or edge location they are targeting.
“We think we have a really good position here with our platforms — giving you this mix-and-match capability but easily be able to scale up and do what our customers want.”
— Bob Ghaffari, Intel
Microsegmentation: Security Through Policy Isolation
Interviewer Justin Warren raised microsegmentation — a topic that was generating significant interest and some frustration among enterprise customers at the time. Microsegmentation divides a network into small, isolated zones so that a breach or exploit in one segment cannot propagate laterally to others. While the security benefits are compelling, operational complexity has been a common barrier to adoption.
Ghaffari acknowledged the complexity but framed microsegmentation as ultimately simplifying the security posture for enterprises that implement it correctly:
- Policy containment: Security and compliance policies are enforced within defined segments rather than relying on a single network perimeter.
- Threat containment: A successful attack or exploit is confined to its originating segment, limiting blast radius.
- Environment clarity: Segmentation can make it easier to reason about what is running where and who has access to what.
The challenge, as Warren noted, is that many customers find initial configuration and ongoing management difficult — which has created a market for third-party tools that layer policy management on top of NSX and competing SDN platforms.
The Intel FPGA N3000: Programmable Acceleration for Virtual Switching
One of the more technically specific topics in the interview was Intel’s FPGA-based network adapter. Ghaffari cited the Intel FPGA N3000 as an example of how customers can go beyond standard NIC capabilities to accelerate custom networking functions and virtual switches in hardware.
FPGA-based SmartNICs occupy a middle ground between fixed-function ASICs and general-purpose CPUs:
- They can be reprogrammed to accelerate specific workloads — including virtual switching, security functions, or 5G baseband processing — without replacing the physical card.
- They offload networking functions from the host CPU, freeing compute resources for application workloads.
- They allow infrastructure operators to adapt hardware behavior as software-defined networking requirements evolve.
For operators running VMware NSX or other SDN overlays at scale, offloading virtual switch processing to a SmartNIC can meaningfully reduce CPU overhead on the host.
Edge Computing: Challenging the “All Intel Does Is Big Xeons” Assumption
Ghaffari described a common misconception among customers evaluating edge deployments: the assumption that Intel hardware is only cost-effective at the high end. Customers planning small-footprint deployments — in retail locations, branch offices, or distributed edge sites — often assume Intel’s price points are prohibitive.
Intel’s response has been to develop add-on processors and purpose-built platforms that hit lower price points while retaining the scalability and software ecosystem compatibility that x86 architecture provides. The implication for DevOps and infrastructure teams is that a consistent programming model — including DPDK-based networking — can be maintained from the core data center out to distributed edge nodes, even at constrained cost targets.
5G and What Comes Next
The full interview also touched on 5G wireless technologies and their intersection with edge computing — a topic that was early-stage in mid-2019 but has since become central to infrastructure planning for carriers and enterprises alike. The core thesis: as 5G shifts more processing to the network edge, the same DPDK-based, software-defined networking capabilities Intel has been building for data centers become relevant for radio access network (RAN) infrastructure.
Key Takeaways for Infrastructure and DevOps Teams
- DPDK remains foundational: If you are running any high-throughput SDN overlay — NSX, OVS-DPDK, or similar — understanding the underlying packet processing library matters for performance tuning.
- SmartNICs are a practical offload option: FPGA-based and ASIC-based SmartNICs are increasingly viable for offloading vSwitch and security functions, particularly in high-density virtualization environments.
- Microsegmentation complexity is real: Plan for operational tooling when deploying microsegmentation at scale; policy management does not end at initial configuration.
- Edge hardware has matured: x86-based edge platforms are no longer exclusively high-cost server-class hardware; purpose-built edge processors have closed the price gap for lightweight deployments.
- Multicloud networking requires a hardware strategy: Consistent connectivity across clouds and edge sites depends on NIC capabilities, driver support, and offload features — not just software configuration.
The original interview was conducted by theCUBE at VMworld, San Francisco. Full interview and transcript available via SiliconANGLE.