Skip to content
Edge Computing · MEC · 5G URLLC · Intel Smart Edge

The Cloud Is 20ms Away.
Your Factory Floor Cannot Wait.

Sending sensor data to a central cloud and waiting for a response kills real-time control. HCT deploys Multi-access Edge Computing (MEC) at URLLC latencies — sub-5ms round-trip — directly at your facility perimeter, powered by Intel Xeon Scalable and Smart Edge Open.

Talk to an MEC ArchitectPrivate 5G + Edge
<5msRound-Trip Latency at Edge
ETSI MECMEC 003 / 011 Standards
10xFaster vs Cloud Round-Trip
IntelSmart Edge + Xeon + OpenVINO
Why Edge Matters

When Latency Kills Operations, the Cloud Is Not an Answer

Central cloud infrastructure sits 20-80ms away from your OT floor. For AGV collision avoidance, SCADA protection switching, and real-time video analytics — that gap is catastrophic.

AGV Collision Avoidance Cannot Tolerate 50ms

Automated Guided Vehicles running at 2 m/s travel 10cm in 50ms — beyond safety stopping distance. IEC 61508 SIL 2 AGV systems require end-to-end control loop under 5ms. Only local edge compute achieves this.
50ms cloud RTT = 10cm AGV overrun at 2m/s speed

IEC 61850 GOOSE Trips on 4ms — Not 40ms

Power grid protection GOOSE messages (IEC 61850-8-1) must trigger circuit breaker trips within 4ms of fault detection. Public 4G/5G delivers 20-50ms. Only local 5G with on-site MEC delivers the sub-4ms required.
GOOSE protection latency requirement: 4ms. Public 4G delivers: 20-50ms

Video Analytics at 4K Needs Local GPU, Not WAN

A single 4K camera stream generates 12-25 Mbps. Running YOLO object detection on 50 cameras in cloud requires 600 Mbps WAN — plus 80ms round-trip. Edge GPU inference processes 50 streams locally in under 30ms.
50-camera site: 600 Mbps WAN vs 0 WAN with edge GPU inference

OT Data Must Never Cross Untrusted WAN

SCADA, DCS, and historian data carries critical infrastructure fingerprints. Sending OT telemetry to a public cloud violates IEC 62443, NERC CIP, and national critical-infrastructure frameworks. Local edge = data stays on-site.
national critical-infrastructure: OT data must not traverse untrusted public networks
MEC Architecture

Three-Layer Edge Compute Stack

HCT deploys ETSI MEC-compliant infrastructure with Intel Xeon compute, Smart Edge orchestration, and OpenVINO AI inference — all co-located with the 5G core UPF for URLLC performance.

Application Layer

🧠 AI / ML Inference Applications

OpenVINO-optimised models: object detection (YOLO), anomaly detection (autoencoder), predictive maintenance (LSTM). Runs on Intel Core Ultra / Xeon E with AMX acceleration. Inference: 15-30ms per frame.

🏭 Industrial OT Applications

SCADA/HMI, AGV fleet manager, digital twin sync, AR guidance server. Containerised on K8s with real-time process scheduling. Direct southbound to PLCs/DCS via OPC-UA, Modbus TCP, or PROFINET.

📹 Video Analytics Engine

50+ concurrent stream processing. License plate recognition (LPR), crowd density, perimeter intrusion detection, hard hat/PPE compliance. MediaSDK hardware decode on Xeon GPU tile.

Platform Layer

⚙️ Intel Smart Edge Open (OpenNESS) Orchestration

Kubernetes-native MEC platform. Manages microservice lifecycle, network policy (Calico CNI), GPU/FPGA resource allocation, and multi-access traffic steering between 5G UPF, Wi-Fi, and wired interfaces. ETSI MEC 003/011 API layer for 3rd-party app onboarding.

🔐 SGX Trusted Execution Environment

Intel SGX enclaves protect sensitive AI models and cryptographic keys from hypervisor-level access. Critical for sovereign deployments where model IP and OT data must be hardware-isolated.

Infrastructure Layer

📡 5G UPF (User Plane Function)

Co-located UPF anchors local data path. Traffic breakout at edge — zero WAN hop for URLLC slices. N6 interface connects UPF directly to local app servers. N4 from SMF controls session routing.

🖥️ Intel Xeon Scalable Compute

2-4 server nodes per edge site. 3rd/4th Gen Xeon with AMX (matrix extensions), DSP offload, and integrated GPU tiles for media. NVMe storage for local historian. 25GbE fabric to 5G RAN.

📶 5G URLLC Radio Layer

Private 5G RAN (SA mode) with URLLC slice: 5ms E2E latency budget, 99.9999% reliability target. Band 78 (3.5GHz) for capacity. Band 28 (700MHz) for deep indoor penetration. RSRP-based handover <1ms.

Industry Applications

Edge Computing Across Critical Sectors

🏭 Smart Manufacturing

AGV fleet control, robot arm coordination, CNC monitoring, AR-guided assembly. Digital twin sync at 10ms update rate. OEE dashboard from 10,000+ OPC-UA sensors.

Target: <5ms AGV control latency

⚡ Power Grid Operations

IEC 61850 GOOSE relay protection, PMU synchrophasor collection, SCADA polling, DER coordination. All on-site — zero WAN dependency for protection functions.

Target: <4ms GOOSE protection trip

🚢 Port Terminal Automation

Crane remote control, AGV collision avoidance, RFID/barcode container scan processing, vessel berthing AI. All latency-sensitive workloads processed pier-side.

Target: <5ms crane actuation

🛢️ Oil and Gas Facilities

ATEX-zone video analytics, gas sensor fusion, drill floor monitoring, SCADA/DCS local processing. Data never leaves the facility boundary — meets operator data sovereignty policies (proven with a national oil company).

Target: <10ms SCADA round-trip

👮 Public Safety Command

AI-powered CCTV analysis at police command posts, ALPR on patrol vehicle footage, facial recognition (authorized use), incident correlation from 50+ camera feeds.

Target: <30ms inference per 4K frame

🏟️ Airports and Large Venues

Passenger flow analytics, baggage carousel monitoring, airside vehicle tracking, gate management AI. Local edge eliminates WAN jitter that causes inconsistent detection results.

Target: <100ms alert generation

Frequently Asked Questions

What is Multi-access Edge Computing (MEC)?

MEC places compute and application workloads at the network edge, on-site or near-site, instead of in a central cloud. This removes the round trip to a distant data centre and is what makes single-digit millisecond application latency achievable.

How much latency does edge computing actually save?

A public cloud round trip typically adds around 20ms or more. Processing at the edge with a local UPF allows HCT Group deployments to achieve under 5ms round trip, which is the threshold required for motion control, machine vision and remote operation.

Does edge computing keep data on-site?

Yes. With a local user plane and on-premises edge compute, operational data is processed and can be retained on-site, which supports both latency requirements and in-country data residency obligations.

What workloads justify edge deployment?

Machine vision and quality inspection, AGV and robotics control, real-time video analytics, augmented reality maintenance, and any closed-loop control where a cloud round trip is too slow or a connectivity outage is unacceptable.

Start with Edge

Bring Compute to Your Operations — Not the Other Way Around

HCT designs, deploys, and manages multi-access edge compute infrastructure for the most latency-sensitive industries in Yemen and the wider region.

Request MEC Design WorkshopIntel Smart Edge Platform