Hybrid Cloud Deployment Guide: AWS, Microsoft Azure & Google Cloud (GCP)
Architecting and Qualifying GxPHistorian Across On-Premises OT (Purdue Levels 1–2.5/3) and Cloud VPC/VNet Environments in Amazon Web Services, Microsoft Azure, and Google Cloud
1. Why Hybrid Cloud (Not 100% Cloud-Only) Is Mandatory in cGMP Manufacturing
In regulated pharmaceutical, biotech, and cell/gene therapy manufacturing (FDA 21 CFR Part 11, EU GMP Annex 11, and ISA-95 / IEC 62443 Purdue Model), attempting to poll cleanroom equipment directly from a public cloud region without an on-premises edge layer fails both engineering physics and GMP data integrity mandates:
- Non-Routable Layer 2 / Localhost Industrial Protocols:
- Cleanroom skids and lab instruments communicate over protocols that cannot (and must never) traverse a WAN router directly to the cloud:
- BACnet/IP (
UDP :47808) relies on local subnet broadcasts (Who-Is/I-Am/SubscribeCOV). - Classic Windows OPC DA 2.05a (used by Cytiva UNICORN™ 7 and legacy Emerson DeltaV™ stations) uses local Microsoft COM (
opcda_windows_forwarder.py). - Siemens S7comm (
TCP :102) and Modbus TCP (TCP :502) lack native transport encryption or user authentication. - RS-232 Serial (
MT-SICSBalances) and Watch-Folder Flat Files (PI UFL) exist on isolated cleanroom VLANs.
- BACnet/IP (
- Cleanroom skids and lab instruments communicate over protocols that cannot (and must never) traverse a WAN router directly to the cloud:
- Zero-Data-Loss Autonomy During WAN Outages (
URS-005/ ALCOA+ Contemporaneous & Complete):- A 14-day perfusion bioreactor campaign or a 72-hour lyophilization cycle cannot lose GMP batch records if a backhoe severs the facility’s fiber-optic ISP link or a cloud VPN tunnel re-keys.
- Therefore, Pillar 1 (
edge-collector) ALWAYS runs on-premises at Purdue Level 2 / 2.5, normalizing all field protocols into the canonical 6-field tuple(tag_id, ts, val, str_val, quality, status_code), executing Stage 1 Exception/Deadband (ExcDev/ExcMax) and optional Stage 2 Swinging Door (CompDev/CompMax) filtering in memory, and persisting up to 100 GB (>14 days for 25,000 tags) in its local disk-backed Write-Ahead Log (/var/lib/telegraf/buffer).
Because GxPHistorian cleanly decouples Pillar 1 (Edge Collector) from Pillar 2 (ClickHouse Archive), Pillar 3 (PostgreSQL 16 + FastAPI + MCP Control Plane), and Pillar 4 (Grafana HMI, Excel DataLink & Parquet Lakehouse), organizations can choose the exact hybrid boundary between their plant floor and AWS, Microsoft Azure, or Google Cloud (GCP).
2. Three Selectable Hybrid Cloud Topologies
flowchart LR
subgraph Plant["On-Premises cGMP Plant (Purdue Levels 1 - 2.5)"]
SKIDS["Cleanroom Instruments\nOPC UA · OPC DA · BACnet\nModbus · S7 · RS-232 · UFL"]
EDGE["Pillar 1: edge-collector\nStage 1 ExcDev + Stage 2 SDT\n100 GB Local Disk WAL (>14d Buffer)"]
SKIDS --> EDGE
end
subgraph Choices["User-Selectable Hybrid Cloud Split"]
PA["Pattern A: Thick Plant Core + Cloud WORM & Lakehouse\n• On-Prem (L3): ClickHouse + Postgres + API + Grafana\n• Cloud (L4): 15-Yr WORM S3/Blob/GCS + Athena/Fabric/BigQuery"]
PB["Pattern B: Thin Plant Edge + Cloud Historian Core\n• On-Prem (L2.5): edge-collector IPC only\n• Cloud (VPC/VNet): ClickHouse + RDS/FlexibleServer/CloudSQL + API + Grafana"]
PC["Pattern C: Multi-Site Hub & Spoke Federation\n• On-Prem (Per Site): Autonomous Pillars 1-4\n• Cloud (Global Hub): Dual-Write ClickHouse + Cross-Site Golden Batch & AI"]
end
EDGE ==>|"Outbound HTTPS / mTLS (:8443)\nDirect Connect / ExpressRoute / Interconnect"| Choices| Hybrid Pattern | On-Premises Footprint (Plant OT) | Cloud Footprint (AWS / Azure / GCP) | WAN Outage Behavior & Ideal Use Case |
|---|---|---|---|
| Pattern A: Thick Plant Core + Cloud 15-Year WORM & Lakehouse (Recommended for Commercial cGMP Plants) | All 4 Pillars (edge-collector, clickhouse-archive 10-yr hot NVMe, postgres-control, historian-api, gxphistorian-mcp, grafana-vision) run in Purdue Level 2.5 / Level 3 on-premises. | • 15-Year WORM Cold Tier (ClickHouse tiered volume + pgBackRest PITR backups)• Zero-ETL Parquet Lakehouse ( Athena, Fabric, BigQuery) | 100% WAN-independent plant operation. Cleanroom HMIs, Event Frame detection, and §11.200 e-signatures work with zero internet/WAN connectivity. Cloud provides elastic 15-year 21 CFR 211.180 WORM retention and global MSAT analytics. |
| Pattern B: Thin Plant Edge + Cloud Historian Core (Recommended for Clinical/Pilot Biotechs & Distributed Sites) | Pillar 1 (edge-collector) only on dual fanless DIN-rail IPCs (deploy/hybrid-cloud/docker-compose.edge-only.yml) with a 100 GB local Disk WAL + Windows OPC DA sidecars. | Pillars 2, 3 & 4 (clickhouse-archive or ClickHouse Cloud, Managed PostgreSQL 16, historian-api, gxphistorian-mcp, and grafana-vision) inside private cloud VPC/VNet. | Zero database servers to maintain on-site. If WAN drops, the Level 2.5 DIN-rail IPC buffers of telemetry on local pSLC NVMe and flushes chronologically when WAN recovers. |
| Pattern C: Multi-Site Hub & Spoke Federation (Recommended for Global Multi-Site Enterprises & CDMOs) | Full Pillars 1–4 at each manufacturing site (e.g., Fremont, Boston, Dublin) for local batch control + Telegraf dual-write (deploy/hybrid-cloud/telegraf-hybrid-dualwrite.conf). | Global Enterprise Hub (Pillars 2–4) aggregating all sites for cross-site Golden Batch overlays (/api/v1/batches/compare) and global AI investigations. | Best of both worlds: Every plant is 100% locally autonomous, while global MSAT and QA teams query a unified multi-site cloud historian and Lakehouse. |
3. Master Cloud Service Mapping Matrix (AWS vs. Azure vs. Google Cloud)
GxPHistorian is engineered without proprietary database lock-in:
- Pillar 2 (
clickhouse-archive) uses standard ClickHouse 24.8 LTS (Apache-2.0)MergeTreeandAggregatingMergeTreetables with built-in<type>s3</type>and<type>azure_blob_storage</type>virtual filesystems. - Pillar 3 (
postgres-control) uses standard PostgreSQL 16contribextensions (ltree,btree_gist,pgcrypto) and pure PL/pgSQL WORM triggers (fn_enforce_part11_audit(),fn_prevent_audit_tampering()) that run unmodified on managed cloud PostgreSQL services without requiring superuser OS filesystem access.
| GxPHistorian Subsystem | Amazon Web Services (AWS) | Microsoft Azure | Google Cloud (GCP) |
|---|---|---|---|
Pillar 1: Level 2.5 Edge Runtime (edge-collector) | Fanless DIN-Rail IPC (Docker), AWS IoT Greengrass v2, or AWS Outposts 1U/2U rack server | Fanless DIN-Rail IPC (Docker), Azure IoT Operations, or Azure Local (Stack HCI) | Fanless DIN-Rail IPC (Docker) or Google Distributed Cloud (GDC) Edge |
Private OT-to-Cloud Transport (21 CFR §11.30) | AWS Site-to-Site VPN / AWS Direct Connect + AWS PrivateLink (:8443 mTLS) | Azure VPN Gateway / Azure ExpressRoute + Azure Private Link (:8443 mTLS) | Cloud HA VPN / Dedicated Interconnect + Private Service Connect (:8443 mTLS) |
Pillar 2: clickhouse-archive Hot Tier | • Self-Managed: EC2 i4i.4xlarge (Nitro NVMe) or m7i.4xlarge + EBS gp3/io2• Managed: ClickHouse Cloud on AWS | • Self-Managed: Azure VM Standard_L16s_v3 (NVMe) or Standard_E16bs_v5 + Premium SSD v2• Managed: ClickHouse Cloud on Azure | • Self-Managed: GCE n2-standard-16 (Local SSD) or c3-standard-22 + Hyperdisk Extreme• Managed: ClickHouse Cloud on GCP |
Pillar 2 & 3: 15-Year WORM Cold Storage (21 CFR 211.180) | Amazon S3 Object Lock (COMPLIANCE Mode, 15-Yr Retention) + AWS KMS CMK | Azure Immutable Blob Storage (Locked Time-Based Retention Policy) + Azure Key Vault | Google Cloud Storage Bucket Lock (Locked Retention Policy, 15-Yr) + Cloud KMS |
Pillar 3: postgres-control Managed Relational & WORM Audit | Amazon RDS for PostgreSQL 16 (Multi-AZ) or Amazon Aurora PostgreSQL (ltree, btree_gist, pgcrypto) | Azure Database for PostgreSQL — Flexible Server (Zone-Redundant HA, ltree, btree_gist, pgcrypto) | Cloud SQL for PostgreSQL 16 (Enterprise Plus HA) or AlloyDB for PostgreSQL (ltree, btree_gist, pgcrypto) |
Pillar 3: historian-api & gxphistorian-mcp Compute | Amazon ECS on AWS Fargate or Amazon EKS behind Internal ALB | Azure Container Apps (VNet-Injected) or Azure Kubernetes Service (AKS) | Google Cloud Run (Internal VPC Ingress) or GKE Autopilot |
Pillar 3: Private GxP AI Model (gxp_ai_agent.py) | Amazon Bedrock (anthropic.claude-sonnet-4 / amazon.nova-pro via VPC Endpoint) | Azure OpenAI Service (gpt-4o via Private Link) | Google Cloud Vertex AI (gemini-2.5-flash / gemini-2.5-pro via Private Service Connect) |
Pillar 4: grafana-vision & Zero-ETL Lakehouse | Amazon Managed Grafana + AWS Glue Data Catalog + Amazon Athena / Redshift Spectrum | Azure Managed Grafana (Entra ID) + Microsoft Fabric OneLake / Azure Synapse / Databricks | Grafana on Cloud Run behind Cloud IAP + BigQuery BigLake External Parquet Tables |
4. Amazon Web Services (AWS) Hybrid Implementation Guide
Step 1: Provision the 15-Year WORM S3 Bucket (COMPLIANCE Mode)
Create an Amazon S3 bucket with S3 Object Lock enabled in COMPLIANCE mode (which prevents even the AWS Account Root user from deleting or overwriting archived telemetry or pgBackRest audit logs before the 15-year retention period expires):
aws s3api create-bucket \ --bucket biocorp-gxphistorian-worm-15yr \ --region us-east-1 \ --object-lock-enabled-for-bucket
aws s3api put-object-lock-configuration \ --bucket biocorp-gxphistorian-worm-15yr \ --object-lock-configuration '{ "ObjectLockEnabled": "Enabled", "Rule": { "DefaultRetention": { "Mode": "COMPLIANCE", "Years": 15 } } }'Mount deploy/hybrid-cloud/clickhouse-tiered-storage/aws_s3_worm.xml to /etc/clickhouse-server/config.d/tiered_storage.xml on your ClickHouse instance (on-premises in Pattern A or EC2 in Pattern B/C):
<clickhouse> <storage_configuration> <disks> <hot_nvme> <path>/var/lib/clickhouse/</path> <keep_free_space_bytes>1099511627776</keep_free_space_bytes> </hot_nvme> <cold_worm_s3> <type>s3</type> <endpoint>https://s3.us-east-1.amazonaws.com/biocorp-gxphistorian-worm-15yr/telemetry/</endpoint> <use_environment_credentials>true</use_environment_credentials> <metadata_path>/var/lib/clickhouse/disks/cold_worm_s3/</metadata_path> <cache_enabled>true</cache_enabled> <data_cache_max_size>107374182400</data_cache_max_size> </cold_worm_s3> </disks> <policies> <gxp_10yr_nvme_15yr_worm> <volumes> <hot_volume><disk>hot_nvme</disk></hot_volume> <cold_volume><disk>cold_worm_s3</disk></cold_volume> </volumes> <move_factor>0.2</move_factor> </gxp_10yr_nvme_15yr_worm> </policies> </storage_configuration></clickhouse>Step 2: Configure Amazon RDS for PostgreSQL 16 & ECS Fargate (Pattern B / C)
- Provision a Multi-AZ Amazon RDS for PostgreSQL 16 (
db.m7g.xlargeordb.r7g.xlargewithgp3Encrypted Storage) in an isolated database subnet. - Apply
init-postgres/01_control_plane_part11.sql(ltree,btree_gist, andpgcryptoare pre-installed in RDS PostgreSQL 16). - Configure
historian-apiandgxphistorian-mcpon Amazon ECS Fargate with TLS enforcement and Amazon Bedrock private AI inference:
PG_HOST=gxp-postgres.cluster-xxxx.us-east-1.rds.amazonaws.comPG_PORT_INTERNAL=5432PG_SSLMODE=verify-fullCH_HOST=clickhouse-internal.biocorp.aws.internalCH_PORT_INTERNAL=8443CH_SECURE=trueGXPHISTORIAN_AI_MODEL=bedrock:anthropic.claude-3-5-sonnet-20241022-v2:0GXPHISTORIAN_USE_LIVE_LLM=1Step 3: Query Snappy Parquet Exports in Amazon Athena
When POST /api/v1/lakehouse/export-parquet writes Snappy-compressed Parquet files to S3, register the external table in AWS Glue / Amazon Athena for zero-ETL MSAT analytics:
CREATE EXTERNAL TABLE IF NOT EXISTS gxp_lakehouse.sensor_telemetry ( tag_id INT, tag_name STRING, asset_path STRING, eng_units STRING, ts TIMESTAMP, val DOUBLE, str_val STRING, quality INT)STORED AS PARQUETLOCATION 's3://biocorp-gxphistorian-worm-15yr/lakehouse-parquet/';5. Microsoft Azure Hybrid Implementation Guide
Step 1: Provision Azure Immutable Blob Storage (Locked 15-Year Policy)
Create an Azure Storage Account with Version-Level Immutability Support and lock a 15-year (5,478-day) time-based retention policy:
az storage account create \ --name biocorpgxpworm15yr \ --resource-group rg-gxp-historian-prod \ --location eastus2 \ --sku Standard_ZRS \ --kind StorageV2 \ --enable-hierarchical-namespace false
az storage container create \ --account-name biocorpgxpworm15yr \ --name gxphistorian-telemetry-worm \ --enable-vlw true
az storage container immutability-policy create \ --account-name biocorpgxpworm15yr \ --container-name gxphistorian-telemetry-worm \ --period 5478 \ --allow-protected-append-writes trueMount deploy/hybrid-cloud/clickhouse-tiered-storage/azure_blob_worm.xml on your ClickHouse node to tier partitions older than 10 years directly to Azure Immutable Blob Storage using <type>azure_blob_storage</type> and Azure Managed Identity (<use_managed_identity>true</use_managed_identity>).
Step 2: Configure Azure Database for PostgreSQL — Flexible Server (Pattern B / C)
Allow-list the three standard PostgreSQL extensions required by GxPHistorian (ltree, btree_gist, pgcrypto) on your Zone-Redundant Azure Database for PostgreSQL Flexible Server:
az postgres flexible-server parameter set \ --resource-group rg-gxp-historian-prod \ --server-name gxp-postgres-prod \ --name azure.extensions \ --value "ltree,btree_gist,pgcrypto"Configure historian-api and gxphistorian-mcp in Azure Container Apps (VNet-injected) or AKS:
PG_HOST=gxp-postgres-prod.postgres.database.azure.comPG_PORT_INTERNAL=5432PG_SSLMODE=verify-fullCH_HOST=clickhouse-internal.biocorp.azure.internalCH_PORT_INTERNAL=8443CH_SECURE=trueGXPHISTORIAN_AI_MODEL=openai:gpt-4oStep 3: Query Parquet Telemetry in Microsoft Fabric OneLake & Azure Synapse
Create an ADLS Gen2 OneLake Shortcut in Microsoft Fabric or query directly via Azure Synapse Serverless SQL Pool:
SELECT tag_name, asset_path, AVG(val) AS avg_val, MIN(val) AS min_val, MAX(val) AS max_valFROM OPENROWSET( BULK 'https://biocorpgxpworm15yr.dfs.core.windows.net/gxphistorian-lakehouse/*.parquet', FORMAT = 'PARQUET') AS telemetryWHERE quality >= 192GROUP BY tag_name, asset_path;6. Google Cloud (GCP) Hybrid Implementation Guide
Step 1: Provision Google Cloud Storage with Bucket Lock (15-Year Locked Policy)
Create a dual-region Google Cloud Storage bucket encrypted with a Cloud KMS Customer-Managed Encryption Key (CMEK) and lock a 15-year (473,364,000 seconds) retention policy:
gcloud storage buckets create gs://biocorp-gxphistorian-worm-15yr \ --location=us \ --uniform-bucket-level-access \ --retention-period=473364000s
# Permanently lock the 15-year WORM retention policy (irreversible per 21 CFR §11.10(c))gcloud storage buckets update gs://biocorp-gxphistorian-worm-15yr \ --lock-retention-periodMount deploy/hybrid-cloud/clickhouse-tiered-storage/gcp_gcs_worm.xml on your ClickHouse node to tier historical partitions to GCS via Google Cloud Storage’s S3-interoperable XML endpoint (https://storage.googleapis.com/biocorp-gxphistorian-worm-15yr/telemetry/).
Step 2: Configure Cloud SQL for PostgreSQL 16 & Cloud Run / GKE (Pattern B / C)
- Provision a Regional HA Cloud SQL for PostgreSQL 16 (Enterprise Plus edition) instance with a Private Service Connect IP (
10.128.0.15) and applyinit-postgres/01_control_plane_part11.sql. - Deploy
historian-apiandgxphistorian-mcpon Google Cloud Run (with Direct VPC Egress and--ingress=internal) wired to Vertex AI (Gemini 2.5):
PG_HOST=10.128.0.15PG_PORT_INTERNAL=5432PG_SSLMODE=verify-caCH_HOST=clickhouse-internal.biocorp.gcp.internalCH_PORT_INTERNAL=8443CH_SECURE=trueGXPHISTORIAN_AI_MODEL=google-vertex:gemini-2.5-flashGXPHISTORIAN_USE_LIVE_LLM=1Step 3: Register BigQuery BigLake External Parquet Table
Expose exported Snappy Parquet files in BigQuery BigLake for enterprise SQL and Vertex AI Workbench Golden Batch modeling:
CREATE EXTERNAL TABLE IF NOT EXISTS `biocorp-gxp.lakehouse.sensor_telemetry`WITH CONNECTION `biocorp-gxp.us.biglake-gcs-conn`OPTIONS ( format = 'PARQUET', uris = ['gs://biocorp-gxphistorian-worm-15yr/lakehouse-parquet/*.parquet']);7. Deploying On-Premises Edge Collectors (Pattern B & Pattern C)
A. Pattern B: Thin Edge Collector Stack (deploy/hybrid-cloud/docker-compose.edge-only.yml)
On your Purdue Level 2.5 fanless DIN-rail Industrial PC (e.g., Advantech UNO-2484G or OnLogic Karbon 800), run only Pillar 1 (edge-collector) with a 100 GB local NVMe WAL pointing outbound over HTTPS (:8443) to your AWS, Azure, or GCP ClickHouse endpoint:
export CH_USER="historian_admin"export CH_PASSWORD="Your_Cloud_ClickHouse_Password!"export CH_INGEST_URL="https://clickhouse-internal.biocorp.cloud.internal:8443/?query=INSERT%20INTO%20pi_historian.sensor_data%20FORMAT%20JSONEachRow"
docker compose -f deploy/hybrid-cloud/docker-compose.edge-only.yml up -dB. Pattern C: Multi-Site Hub & Spoke Dual-Write (deploy/hybrid-cloud/telegraf-hybrid-dualwrite.conf)
For multi-site enterprises running both a local Purdue Level 3 plant historian and a global cloud aggregator, mount deploy/hybrid-cloud/telegraf-hybrid-dualwrite.conf into edge-collector. Telegraf maintains two independent [[outputs.http]] queues (onprem_level3_clickhouse and cloud_enterprise_clickhouse_hub) so a cloud WAN maintenance window never delays local cleanroom data ingestion.
C. Programmatic Blueprint & Readiness API (GET /api/v1/cloud/hybrid-blueprints)
You can query historian-api at runtime to retrieve validated ClickHouse WORM XML policies, environment profiles, and Lakehouse DDL for any cloud provider (aws, azure, gcp) and topology (pattern_a_worm_lakehouse, pattern_b_cloud_core, pattern_c_hub_spoke):
curl -s "http://localhost:8020/api/v1/cloud/hybrid-blueprints?provider=aws&pattern=pattern_b_cloud_core" | jq .8. 21 CFR Part 11, EU Annex 11 & ISPE GAMP 5 Cloud Qualification Checklist
When qualifying a hybrid deployment under ISPE GAMP 5 (Second Edition) and EU GMP Annex 11 (Section 4.3 — Cloud / IT Supplier Management), document the following Shared Responsibility controls in your site Validation Master Plan (VMP):
| Regulatory Requirement | Cloud Provider Responsibility (IaaS / PaaS) | Regulated User / GxPHistorian Technical Control |
|---|---|---|
21 CFR §11.30 — Controls for Open Systems (Authenticity, integrity, and confidentiality of records transmitted over WAN) | Physical fiber security, Direct Connect / ExpressRoute / Interconnect colocation security, and FIPS 140-3 cryptographic modules. | • Outbound-only TLS 1.3 / mutual TLS (mTLS) from Purdue Level 2.5 edge-collector (:8443) and PG_SSLMODE=verify-full.• Zero inbound firewall ports opened into the cleanroom OT network. |
21 CFR §11.10(e) & ALCOA+ — Contemporaneous & Complete Capture During Network Loss | Regional VPC/VNet availability and Multi-AZ load balancer failover. | • Local 100 GB Telegraf Disk WAL (/var/lib/telegraf/buffer) on industrial pSLC NVMe at Purdue Level 2.5.• Buffers of 25,000 tags with original equipment timestamps ( DateTime64(3, 'UTC')). |
21 CFR §11.10(e) — Tamper-Evident WORM Audit Trail | Managed PostgreSQL (RDS, Azure Flexible Server, Cloud SQL) underlying storage durability and automated encrypted backups. | • Database-enforced PL/pgSQL triggers (fn_enforce_part11_audit() and fn_prevent_audit_tampering()) blocking UPDATE/DELETE/TRUNCATE even by DB superusers.• Cryptographic SHA-256 hash chain verified via GET /api/v1/audit-trail/verify. |
21 CFR 211.180 & EU Annex 11 §17 — 15-Year Immutable Batch & Telemetry Archiving | Hardware-enforced object immutability (AWS S3 Object Lock COMPLIANCE, Azure Locked Immutability Policy, GCP Locked Bucket Retention). | • ClickHouse multi-volume storage policy (<move_factor>0.2</move_factor> and TTL ... TO VOLUME 'cold_volume') moving aged partitions to WORM object storage without DELETE. |
EU Annex 11 §4.3 & GAMP 5 — Cloud Supplier Assessment & Data Residency | SOC 1/2/3, ISO/IEC 27001, ISO/IEC 27017/27018, and GxP Qualification packs (AWS GxP, Azure GxP, Google Cloud GxP). | • Pin all cloud resources, WORM buckets, and private AI endpoints (Bedrock, Azure OpenAI, Vertex AI) to approved sovereign/regional jurisdictions (e.g., us-east-1, eu-central-1). |
9. GxPHistorian Cloud Commercial Editions & Transparent Pricing
While the GxPHistorian Open-Source Core (MIT) is 100% free forever for self-managed deployments, regulated biotech and pharmaceutical organizations frequently prefer a turnkey, pre-validated cloud control plane backed by a signed EU GMP Annex 11 §4.3 Quality Technical Agreement (QTA), version-pinned Change-Controlled Release Channels (2026.Q4-LTS), and Unlimited Grafana Vision HMI & Excel DataLink seats ($0 per-seat tax).
Dual-Track Commercial Delivery Model
| Commercial Edition | Architecture & Cloud Residency | Included Capacity | Annual Subscription | Ideal Regulated Buyer |
|---|---|---|---|---|
1. GxPHistorian Cloud — Lab & PD (GXP-CLOUD-LAB-2K5) | Managed Single-Tenant Cloud (Pattern B): Dedicated isolated VPC managed by Saram Consulting + 1 Purdue L2.5 Edge Collector | Up to 2,500 active tags · 1 Site · Unlimited Users · 3-Yr Hot NVMe + 15-Yr WORM Archive · Automated IQ/OQ Sandbox Binder | $18,000 / yr ($1,500 / mo) | Process Development (PD), Analytical Development, and Pre-Clinical Biotech Labs |
2. GxPHistorian Cloud — Clinical & Pilot cGMP (GXP-CLOUD-CLINICAL-25K) | Managed Single-Tenant Cloud (Pattern B): Dedicated Multi-AZ VPC + 2 Redundant Purdue L2.5 Edge Collectors (>14-day Disk WAL) | Up to 25,000 active tags · 1 Site · Unlimited Users · 10-Yr Hot NVMe + 15-Yr WORM · Okta / Entra ID §11.200 SSO · Signed EU Annex 11 §4.3 QTA & Change-Controlled IQ/OQ Releases | $54,000 / yr ($4,500 / mo) | Phase I–III Clinical Manufacturing Plants, Cell & Gene Therapy (CAR-T / mRNA) Suites, and Nimble CDMOs |
3. GxPHistorian Cloud — Commercial cGMP & Enterprise Hub (BYOC) (GXP-CLOUD-BYOC-100K) | Bring Your Own Cloud (Pattern B & C): Deployed directly inside the Customer’s AWS / Azure / GCP VPC via Cloud Marketplace with Customer-Managed KMS Keys (CMK/CMEK) | Up to 100,000 active tags (up to 3 Sites) · Unlimited Global Users · Cross-Site Golden Batch Hub · Private Bedrock / Azure OpenAI / Vertex AI Co-Pilot · Zero-ETL Lakehouse · Annual QA Supplier Audit Rights | $150,000 / yr ($12,500 / mo — 100% AWS EDP / Azure MACC / GCP Commit eligible) | Commercial Biomanufacturing Facilities, Mid-to-Large Pharma, and Multi-Site Global CDMOs |
Add-On: Pre-Qualified GxPHistorian Edge Box (GXP-EDGE-BOX-L25) | Plug-and-Play Purdue Level 2.5 Appliance: Fanless DIN-rail IPC (Advantech UNO-2484G or OnLogic Karbon 800, 512 GB pSLC NVMe, Dual Isolated LAN) | Pre-flashed with edge-collector, 100 GB Telegraf Disk WAL (>14-day autonomy), outbound mTLS (:8443), and Factory-Executed FAT/IQ Certificate | $3,500 hardware + $2,500 / yr Fleet Care | Cleanroom suites requiring zero-IT shop-floor commissioning |
10. Cloud Edge Fleet Enrollment & TCO Calculator REST APIs
A. Enroll & Monitor Purdue Level 2.5 Edge Collectors (POST /api/v1/cloud/edge-agents/heartbeat)
Every Purdue Level 2.5 edge-collector node (or pre-qualified GxPHistorian Edge Box) reports its 100 GB Telegraf Disk WAL utilization, active field protocols, and outbound mTLS certificate expiration to the control plane with mandatory 21 CFR Part 11 attribution (X-Actor-Id & X-Change-Reason):
curl -s -X POST "http://localhost:8020/api/v1/cloud/edge-agents/heartbeat" \ -H "Content-Type: application/json" \ -H "X-Actor-Id: operator_01" \ -H "X-Change-Reason: Enroll Purdue L2.5 DIN-Rail Edge Box for USP Suite 201" \ -d '{ "agent_id": "edge-fremont-usp-01", "site_path": "BioCorp.Fremont.USP.Bioreactor_201", "hardware_model": "Advantech UNO-2484G (512GB pSLC NVMe)", "deployment_pattern": "pattern_b_cloud_core", "cloud_provider": "aws", "wal_buffer_used_gb": 1.25, "wal_buffer_capacity_gb": 100.0, "active_tag_count": 25000, "mtls_cert_days_remaining": 84, "wan_link_status": "CONNECTED_MTLS_8443" }' | jq .To inspect all enrolled Edge Collectors and verify URS-005 ( WAN outage autonomy) compliance across your fleet:
curl -s "http://localhost:8020/api/v1/cloud/edge-agents" | jq .B. Programmatic Cloud Sizing & AVEVA PI TCO Savings Calculator (GET /api/v1/cloud/tco-calculator)
Compute 10-year Hot NVMe + 15-year WORM S3/Blob/GCS storage requirements, get an automated SKU recommendation, and compare 1-year and 3-year Total Cost of Ownership (TCO) against legacy per-tag/per-seat historians (AVEVA PI Flex + manual CSV consulting):
curl -s "http://localhost:8020/api/v1/cloud/tco-calculator?active_tags=25000&named_users=50&site_count=1&cloud_provider=aws" | jq .