Leader in ISG Private Cloud Report 2026.

UnitedLayerBlue Version

UPC Container Cloud

Enterprise-
grade
Kubernetes
and
containers,
simplified.

A managed container platform to run, scale and secure containerized workloads across private, public and edge environments, with GPU acceleration and private AI built in.
See the AI architecture
Try the console

1

Control plane for private, public and edge clusters
0 %
Availability
0
Compliance frameworks supported: HIPAA, SOC 2, PCI DSS

24/7

Technical support
containers.unitedlayer / example-company
All clusters

Clusters

Example tenant · illustrative figures

Clusters4
Pods424
GPUs busy6 / 16
Version1.30
Cluster · Detail
Status
Deploy · checkout v2.4
Rolling out to prod-private and prod-awsLive on prod-private and prod-aws · 0 errors
DeployingDone
Upgrade · all clusters to 1.30
Draining nodes one at a time · no downtime4 of 4 clusters on 1.30 · zero downtime
UpgradingDone
AI job · fine-tune support model
Fine-tuning on 4 GPUs in prod-privateFine-tune finished · model served by the vLLM layer
TrainingDone
prod-privateUnited Private Cloud® · 180 pods · 8 GPUs
Healthy
prod-awsAWS · 120 pods
Healthy
analytics-azureAzure · 64 pods
Healthy
stores-edgeEdge · 60 pods
Healthy
checkout v2.4 is live on two clouds
All 4 clusters on 1.30 · zero downtime
Model live on your private LLM endpoint

Workloads

Across all 4 clusters · illustrative figures

Deployments38
Jobs6
Services52
Errors0
Workload · Where
Status
Scale · checkout 24 → 36 pods
Adding pods on prod-private and prod-aws36 pods ready · traffic balanced
ScalingDone
Restart · payments-api
Rolling restart, one pod at a time16 of 16 pods restarted · 0 errors
RestartingDone
checkoutDeployment · prod-private, prod-aws · 24 pods
Running
payments-apiDeployment · prod-private · 16 pods
Running
support-llmvLLM inference · prod-private · 2 GPUs
Running
recommendationsBatch job · analytics-azure · nightly
Done
store-syncDaemonSet · stores-edge · 60 pods
Running
checkout scaled to 36 pods
payments-api restarted, zero downtime

GPU pools

Shared across teams · illustrative figures

Pools3
GPUs16
Busy6
In use38%
Pool · Detail
Busy
Add GPUs · private-gpu 8 → 12
Provisioning 4 GPUs in United Private Cloud®12 GPUs ready in private-gpu
AddingDone
Rebalance · batch to azure-gpu
Moving training jobs off private-gpuBatch on azure-gpu · inference has headroom
MovingDone
private-gpuprod-private · 8 GPUs · fine-tuning and inference
4 / 8
aws-gpuprod-aws · 4 GPUs · burst inference
2 / 4
azure-gpuanalytics-azure · 4 GPUs · batch training
0 / 4
4 more GPUs ready in private-gpu
GPU load balanced across pools

Pipelines

Today · illustrative figures

Runs14
Passed13
Running1
Failed0
Pipeline · Detail
Status
Run · payments-api v1.9
Build → scan → canary on prod-privateCanary passed · promoted to prod-private
RunningPassed
Roll back · checkout to v2.3
Reverting prod-aws to the last good versioncheckout v2.3 live · 0 errors
Rolling backDone
checkout v2.4Build → scan → deploy to two clouds
Passed
support-model fine-tuneTrain → evaluate → serve on vLLM
Passed
nightly image scanAll registries · 0 critical findings
Passed
payments-api v1.9 promoted
Rolled back in under a minute

Policies

Applied to every cluster · illustrative figures

Policies12
Enforced12
Violations0
Last auditToday
Policy · Scope
Status
Audit · all clusters
Checking 12 policies on 4 clusters12 of 12 passing · 0 violations
AuditingPassed
New policy · no public load balancers
Rolling out to all 4 clustersEnforced on 4 of 4 clusters
ApplyingEnforced
Signed images onlyAll clusters · unsigned images are blocked
Enforced
No privileged containersAll clusters · except approved system pods
Enforced
Data stays in regionprod-private · regulated workloads
Enforced
GPU quota per teamAll GPU pools · fair share
Enforced
Audit clean · 0 violations
New policy enforced everywhere
Included
  • 100+ pre-configured solutions
  • NVIDIA GPUs
  • VMware
  • Azure Hyper-V
  • KVM
  • Oracle
  • SAP
  • VMs
  • Containers
  • Serverless
  • ITSM
  • UnityOne AI™
  • Agentic orchestration
  • Carbon measurement
  • UnitedSecure™
  • UnitedConnect®
  • 99.999% availability
  • 24/7 support
UPC AI Cloud architecture

From GPUs and fast networks
to AI agents that run your cloud

A GPU cluster with a vLLM layer serves your private LLM. A Vector DB adds your own knowledge,
an ML Gateway routes each request to the right AI agent, and the agents deliver capabilities like
cost forecasting, threat response and self-healing. Pick a layer to learn more.
UPC Container Cloud

UPC AI Cloud architecture

AI capabilities
Predictive cost forecastingCapacity planningThreat detection and responseAutomated infra provisioningCentralized knowledge base Automated rightsizingAnomaly detectionIdentity and access managementWorkflow automationSelf-healing systems
Vector DB
LLB Cloud
ML
Gateway
AI agents
Network agentSecurity agentPerformance agentVirtualization agentCost agent
Customer private LLM
vLLM layerUPC AI Cloud GPU cluster
Infrastructure
layer
Features
  • Distributed compute
  • Customized security
  • Cluster deployment
  • 99.999% availability
  • Optimized infrastructure
  • High performance
  • Cost efficient
Data model layer x86 architecture
InfiniBandHigh-speed Ethernet · 100 GbpsCXL · Compute Express LinkFiber connect
High-speed computing
Software-defined database
DBDBDBDBDBDB

AI capabilities

Ten AI-driven capabilities run across the platform, from forecasting cost to fixing problems before you notice them.

Cost forecastingCapacity planningThreat responseSelf-healing

AI agents

Five specialist agents: network, security, performance, virtualization and cost. Each one acts on its own area of your cloud.

NetworkSecurityPerformanceVirtualizationCost

ML Gateway

One entry point between your models, your data and the AI agents. Every request goes through it to the right model.

Single entry pointModel routingAccess control

Vector DB and LLB Cloud

Your own knowledge sits next to the gateway, so models and agents answer from your data.

Vector DBLLB Cloud

GPU cluster and vLLM layer

The UPC AI Cloud GPU cluster runs your customer private LLM, served through the vLLM layer.

GPU clustervLLMPrivate LLM

Data model layer

Fast interconnects on x86 architecture move data between compute and storage without bottlenecks.

InfiniBand100 Gbps EthernetCXLFiber connect

High-speed computing

Dense compute nodes for training, inference and the everyday workloads around them.

Distributed computeHigh performance

Software-defined database

Databases defined in software, so they scale and move with the workloads that use them.

Cluster deploymentScales on demand

Infrastructure layer

The foundation under every layer above, built for security, performance and 99.999% availability.

99.999% availabilityCustomized securityCost efficient

Features

Kubernetes without
the heavy lifting

Pick one to see it in action.

Clusters that patch and upgrade themselves

Automated provisioning, scaling and lifecycle management, so your team ships apps instead of running clusters.

  • Automated provisioning
  • Scaling built in
  • Upgrades with zero downtime
ProvisionScaleUpgrade
Kubernetes, managed
Upgrade three clusters. Zero downtime.
EKS · prod0/4 nodes1/4 nodes2/4 nodes3/4 nodes4/4 nodes
node-1 · v1.29node-1 · upgradingnode-1 · v1.30
node-2 · v1.29node-2 · upgradingnode-2 · v1.30
node-3 · v1.29node-3 · upgradingnode-3 · v1.30
node-4 · v1.29node-4 · upgradingnode-4 · v1.30
AKS · prod0/4 nodes1/4 nodes2/4 nodes3/4 nodes4/4 nodes
node-1 · v1.29node-1 · upgradingnode-1 · v1.30
node-2 · v1.29node-2 · upgradingnode-2 · v1.30
node-3 · v1.29node-3 · upgradingnode-3 · v1.30
node-4 · v1.29node-4 · upgradingnode-4 · v1.30
GKE · prod0/4 nodes1/4 nodes2/4 nodes3/4 nodes4/4 nodes
node-1 · v1.29node-1 · upgradingnode-1 · v1.30
node-2 · v1.29node-2 · upgradingnode-2 · v1.30
node-3 · v1.29node-3 · upgradingnode-3 · v1.30
node-4 · v1.29node-4 · upgradingnode-4 · v1.30
Rolling upgrade v1.29 → v1.30 · pods moved before each node restartsApp serving · 0 errors

One control plane for every cloud

Run clusters on United Private Cloud®, AWS, Azure and the edge, and manage them all from one place.

  • Same tooling on every cloud
  • Move workloads where they fit
  • One view of health and cost
Private cloudAWSAzureEdge
Hybrid and multi-cloud
One control plane. Every cloud.
One control plane · policies, deploys, cost
Private cloudUnited Private Cloud®
AWSburst capacity
Azureanalytics
Edgestores
Same policies and tooling on every cloud · checkout bursts to AWS at peak4 clouds · healthy

Secure and compliant by default

Image scanning, policies and access control are built in and applied to every cluster.

  • Signed images only
  • Zero Trust access
  • Audit-ready reporting
HIPAASOC 2PCI DSS
Security and compliance
Every image checked before it runs.
checkout:2.4signed
Passed
payments-api:1.8signed
Passed
debug-tools:latestunsigned
Blocked
support-llm:3.1signed
Passed
✓HIPAA
✓SOC 2
✓PCI DSS
Policy: signed images only · enforced on every cluster0 violations running

GPU clusters for AI and HPC

Train, fine-tune and serve models on GPU clusters, with a vLLM layer for private LLM endpoints.

  • GPU pools shared across teams
  • vLLM model serving
  • High-speed InfiniBand networking
GPUsvLLMInfiniBand
AI and HPC
Fine-tune, then serve. On your GPUs.
GPU 1GPU 2GPU 3GPU 4GPU 5GPU 6GPU 7GPU 8GPU 9GPU 10GPU 11GPU 12GPU 13GPU 14GPU 15GPU 16
Queue
Fine-tune
Evaluate
Serve on vLLM
private-gpu pool · 6 of 16 GPUs busy · InfiniBandEndpoint live
What’s included

Four services, one platform

KaaS
Fully managed Kubernetes
  • Automated setup, patching and updates
  • Enterprise-grade clusters
  • Built-in monitoring and scaling
  • CI/CD integration
Multi-cloud
Container orchestration
  • Workload portability
  • Security and compliance policies enforced
  • Centralized visibility
  • AWS, Azure, Google Cloud and private clouds
Edge and hybrid
Clusters near your data
  • Edge-optimized clusters
  • Low-latency processing
  • Secure edge, cloud and on-prem links
  • AI-driven workload placement
AI and GPU
Built for AI workloads
  • GPU-powered containers
  • Low-latency compute
  • Auto-scaling for AI jobs
  • Integrated MLOps tooling

Plan a cluster

Sketch your container
platform in a minute

Pick what you run and where. See the cluster layout we would start from.
Workloads
Where they run
Compliance
Clusters
Private cluster on United Private Cloud®
Public cloud cluster (AWS, Azure or Google Cloud)
Edge clusters at your sites
Pick where your workloads run.
Node pools
General pool · web and API containers
Data pool · memory-heavy nodes for analytics
GPU inference pool · GPUs serving models through the vLLM layer
GPU training pool · scales up for jobs, back down after
Pick at least one workload.
Built in
  • One control plane
  • Managed upgrades, monitoring and CI/CD
  • Private LLM, Vector DB and AI agents from the UPC AI Cloud
  • SOC 2HIPAAPCI DSS policies enforced on every cluster

A starting sketch, not a quote. Your UnitedLayer architect sizes the real clusters with you.

See it in motion

Watch containers run themselves

Six moments from UPC Container Cloud, from a zero-downtime upgrade to GPUs busy with AI.
Kubernetes, managed
Upgrade three clusters. Zero downtime.
EKS · prod0/4 nodes1/4 nodes2/4 nodes3/4 nodes4/4 nodes
node-1 · v1.29node-1 · drainingnode-1 · v1.30
node-2 · v1.29node-2 · drainingnode-2 · v1.30
node-3 · v1.29node-3 · drainingnode-3 · v1.30
node-4 · v1.29node-4 · drainingnode-4 · v1.30
AKS · prod0/4 nodes1/4 nodes2/4 nodes3/4 nodes4/4 nodes
node-1 · v1.29node-1 · drainingnode-1 · v1.30
node-2 · v1.29node-2 · drainingnode-2 · v1.30
node-3 · v1.29node-3 · drainingnode-3 · v1.30
node-4 · v1.29node-4 · drainingnode-4 · v1.30
GKE · prod0/4 nodes1/4 nodes2/4 nodes3/4 nodes4/4 nodes
node-1 · v1.29node-1 · drainingnode-1 · v1.30
node-2 · v1.29node-2 · drainingnode-2 · v1.30
node-3 · v1.29node-3 · drainingnode-3 · v1.30
node-4 · v1.29node-4 · drainingnode-4 · v1.30
Rolling upgrade v1.29 → v1.30 · pods moved before each node restartsApp serving · 0 errors
One control plane
Four clusters. One view.
ClusterWherePodsStatus
prod-privateUnited Private Cloud®180 podsHealthy
prod-awsAWS120 podsHealthy
analytics-azureAzure64 podsHealthy
stores-edgeEdge · 12 sites60 podsHealthy
Same deploy, policy and alert rules pushed to all four
One console for private, public and edge · 424 podsAll clusters healthy
GPU autoscaling
GPUs scale with the queue.
GPU nodes4 GPUs8 GPUs12 GPUs4 GPUs
Training queue0 jobs queued3 jobs queued5 jobs runningJobs done
private-gpu pool · scales up for jobs, back down afterPay for what runs
Model delivery
Ship a new model with a canary.
TrainEvaluateCanaryFull rollout
support-model v1v2
v2 · 0% of trafficv2 · 10% canaryv2 · 50% of trafficv2 · 100% · v1 retired
Served by the vLLM layer · automatic rollback if checks failError rate 0.0%
Observability
From app to GPU, one view.
p95 latency182 ms
Errors0.02%
GPU use71%
AppsKubernetesNodesGPUs
Metrics, logs and traces from every layer, in one place4 of 4 layers reporting
Policy as code
Checked on every deploy.
policies.yamlpolicy: signed-images-only applies_to: all-clusters action: blockpolicy: data-stays-in-region applies_to: prod-private
checkout:2.4signedAllowed
debug-tools:latestunsignedBlocked
payments-api:1.9signedAllowed
Same rules on every cluster · every decision logged0 violations running
Live · auto-playingPausedStep 1 of 6Step 2 of 6Step 3 of 6Step 4 of 6Step 5 of 6Step 6 of 6
Observability solution suite

See every container, app and GPU

Monitoring for Kubernetes, Docker and Amazon ECS, with five views across your estate. Pick one.

Every request traced

Application tracing, latency tracking, error detection and user experience insights.

  • Tracing
  • Latency
  • Errors
  • User experience

Every node, disk and GPU watched

Health and capacity for servers, storage, containers and GPUs, at the edge and in the cloud.

  • CPU and memory
  • Storage
  • GPUs
  • Containers

Slow links found fast

Latency, packet loss and throughput between sites, clouds and users, with the path shown end to end.

  • Latency
  • Packet loss
  • Throughput
  • Path tracing

Problems fixed before users notice

AI links related alerts, finds the root cause and can act on it automatically.

  • Alert correlation
  • Root cause
  • Anomaly detection
  • Auto-remediation

Threats seen in context

Suspicious activity, access and configuration changes, shown next to the performance data they affect.

  • Threat detection
  • Access monitoring
  • Config drift
  • Audit trail
Intelligent edge monitoring
See every layer. Fix issues before users feel them.
Infra observability
Network performance
App observability
AIOps
Security observability
Service map · AI-powered Edge site, store 14
Cameras GPU node Inference API Edge storage Store app
● All services healthy · watching 5 services ▲ Slow GPU node spotted · AI is finding the cause ✓ Fixed in minutes · response times back to normal
Store 14 · all layers reporting Root cause: GPU node overloaded · moving inference load Workload rebalanced before users noticed
Monitors Kubernetes · full lifecycle Docker · auto-discovers short-lived containers Amazon ECS and EKS · preset alert thresholds

How it works

From first cluster to
private AI in five steps

Assess

We map your apps, infrastructure and costs, and find what to move first.

OutcomeA migration plan with clear priorities

Build

Managed clusters are set up on private, public and edge environments under one control plane.

OutcomeClusters ready, policies in place

Migrate

Apps move in waves, with infrastructure as code and a test before every cutover.

OutcomeApps live, zero-downtime cutover

Add AI

GPU node pools, a private LLM endpoint and AI agents are added to the same clusters.

OutcomeAI running next to your apps

Operate

UnitedLayer runs the platform day to day: upgrades, monitoring, cost and governance.

OutcomeSteady, governed operations
UnityOne AI™
CMPSecurityK8IaCGovernance
VMs discovered412
Apps mapped38
Run cost / month$86K
Migration readiness · by app
checkout12 VMs · statelessContainerize
search6 VMs · ElasticContainerize
payments8 VMs · PCI scopeRefactor first
legacy-erp22 VMs · licensedKeep as is
UnityOne AI™
CMPSecurityK8IaCGovernance
EKS · prodAWS · us-west-2

6 nodes · 23 pods

CPU 69% · scaled out

AKS · prodAzure · West US

5 nodes · 18 pods

CPU 54% · pod restarted

GKE · analyticsGoogle Cloud · us-central1

4 nodes · 12 pods

CPU 41%

Cost by namespace · this month ↑ EKS checkout busy · scaled 16 → 24 pods
checkout$2.1K
search$1.4K
analytics$0.9K
platform$0.7K
UnityOne AI™
CMPSecurityK8IaCGovernance
migrate.tf · plan+ 3 clusters + 14 node pools + 9 policiesPlan: 26 to add, 0 to change, 0 to destroy✓ Applied in test · ready for wave 2
Migration waves
Wave 1 · checkout, searchCut over · 0 min downtime
Live
Wave 2 · paymentsMoving data · test cutover next
In progress
Wave 3 · analyticsStarts after wave 2
Queued
UnityOne AI™
CMPSecurityK8IaCGovernance
GPU pool · prod-privateUnited Private Cloud® · 8 GPUs

6 of 8 GPUs busy · fine-tune running

Private LLM endpointvLLM layer · live
1,840tokens / sec

p95 latency 210 ms

AI agents on the cluster
Network agentSecurity agentPerformance agentCost agent
UnityOne AI™
CMPSecurityK8IaCGovernance
Policy compliance98%
Alerts auto-resolved41
Config drift0
This month
Budget used
82% · on track
Resources tagged
100%
Clusters on latest version
4 of 4
Patches applied
96% · 2 scheduled

Side by side

Three ways to run containers

Area Self-managed Kubernetes One cloud’s managed service UPC Container Cloud
Operations Your team patches and upgrades Provider runs the control plane Fully managed: setup, patching and updates
Where it runs Where you build it That provider’s regions Private, public and edge, one control plane
AI and GPUs Hardware you buy and wire up Shared GPU capacity GPU cluster, vLLM layer and private LLM
Operations AI Scripts Separate services AI agents for network, security, performance, virtualization and cost
Compliance You build the controls Shared responsibility HIPAA, SOC 2 and PCI DSS support, policies enforced
Visibility Many tools Per provider One observability suite for apps, infra, network, AI and security
Questions

What teams ask before they start

What is UPC Container Cloud?

closed Icon Opend Icon

A managed container platform from UnitedLayer to run, scale and secure containerized workloads across hybrid and multi-cloud environments, from one control plane.

Is Kubernetes fully managed?

closed Icon Opend Icon

Yes. Provisioning, scaling, patching, updates and lifecycle management are automated, with built-in monitoring and CI/CD integration.

Which clouds does it work with?

closed Icon Opend Icon

Private clouds, AWS, Azure and Google Cloud, plus edge locations, with workload portability between them.

How does it support AI?

closed Icon Opend Icon

GPU-powered containers, auto-scaling for AI workloads and integrated MLOps tooling. The UPC AI Cloud architecture adds a GPU cluster with a vLLM layer for your private LLM, a Vector DB, an ML Gateway and AI agents.

Which compliance standards are supported?

closed Icon Opend Icon

HIPAA, SOC 2 and PCI DSS, with encrypted networking and identity-based access control.

What can you monitor?

closed Icon Opend Icon

Kubernetes, Docker and Amazon ECS and EKS, with app, infrastructure, network, AIOps and security observability.

↑