The infrastructure

The Epic infrastructure, layer by layer

An Epic environment is not one system. A clinical click travels through five layers, from the device at the bedside to the operational database. Two more sit beside that path: analytics, which works from a copy of the data, and the foundation that carries everything. This page walks through all seven: what runs there, what a failure looks like on the ward, and which expertise is needed.

This is a general overview of common architecture patterns. The layers show dependencies, not a network trace: every click relies on them, but not every request passes through every service. Every organization’s design differs, and Epic’s own documentation and guidance remain authoritative. Product names describe typical setups and are not recommendations.

Endpoint & peripherals

Where clinicians actually work: thin clients and workstations, workstations on wheels and the devices attached to them. Epic’s mobile apps extend this layer: with Rover, nurses scan wristbands and medications with a phone camera, while Haiku on the phone and Canto on the iPad bring the chart to physicians on the move. The apps connect through Interconnect and depend on device management and full Wi-Fi coverage.

Expertise needed

Endpoint and peripheral integration for Epic
Device configuration, drivers and USB or serial redirection
Print routing to the right device at the right location
Mobile device management for shared and personal devices
Close collaboration with clinical and workplace teams

Lead rolesECSA, Identity & Workplace

What runs here

Thin clients & PCs
Workstations on wheels (WoW)
Badge readers
Label & wristband printers
Barcode scanners
Signature pads
Device drivers & redirection
Smartphones & tablets: Rover, Haiku, Canto

When it fails

A scanner that does not register a barcode, or a label printer that prints at the wrong station, stops medication administration and specimen collection directly at the bedside.

Identity & access

Clinicians share workstations and switch between them all day. Fast, secure user switching is a clinical requirement, not a convenience.

Expertise needed

Directory services and group policy design
SSO integration with Epic and with Citrix sessions
Authentication across multiple session hops
Balancing security policy with clinical workflow

Lead rolesIdentity & Workplace

What runs here

Active Directory
Single sign-on (badge tap)
Multi-factor authentication
Group policies
Role and security models
Certificates

When it fails

If sign-on fails or is slow, clinicians wait at every workstation. Workarounds such as shared accounts appear quickly and become a security and audit problem.

Application delivery

Hyperdrive is Epic’s standard end-user client today. Most organizations replaced the classic Hyperspace client with it in 2022 and 2023; Hyperspace remains the name of the application itself. The key design decision is how Hyperdrive reaches the workstation: installed locally on the endpoint, or delivered through Citrix Virtual Apps and Desktops. Many hospitals combine both. When Hyperdrive is published through Citrix and launched from inside a virtual desktop, double-hop sessions and single sign-on across hops come into play.

Expertise needed

Citrix architecture, image management and capacity planning
Hyperdrive deployment: local install, Citrix or both
Session performance analysis from logon to working context
Coordinating client releases with ECSA and the platform team

Lead rolesApplication Delivery

What runs here

Hyperdrive
Citrix Virtual Apps and Desktops
StoreFront & Delivery Controllers
Session hosts (VDAs)
Citrix Provisioning (PVS)
NetScaler Gateway
Local client deployment
Profile management

When it fails

Slow logons, frozen sessions, a broken image or a faulty client package hit every user who depends on them at once. Session launch time is clinical time.

Web & integration services

Hyperdrive is Epic’s own browser for Hyperspace Web. Hyperspace Web runs on load-balanced web servers and handles the application logic of each click before the data reaches the database. Around it, further Windows and web servers connect Epic with patients, mobile devices, documents, printers and other systems. Interconnect carries Epic’s web services, including those for the mobile apps, for FHIR and other APIs, and for Care Everywhere, which exchanges patient records between organizations using Epic and other interoperable EHRs. Bridges is Epic’s interface module for message-based interfaces such as HL7 v2. Most hospitals add a separate integration engine between Bridges and the many departmental, regional and national systems.

Expertise needed

Windows Server and IIS administration
Load balancing, TLS certificates, DNS and secure external access
Integration engine operations: high availability, queue monitoring, message replay
Epic print infrastructure

Lead rolesECSA

What runs here

Hyperspace Web (HSWeb)
Interconnect web services
Bridges interfaces
Integration engine
Care Everywhere
MyChart patient portal
Web BLOB document storage
Epic Print Service
IIS, load balancers & certificates

When it fails

Depending on which service fails, clinicians cannot work in Hyperspace, orders and results stop flowing, documents do not open, mobile apps or the patient portal are unreachable, or nothing prints. Messages pile up in queues and must be processed in the right order after recovery.

Operational database

At the core is Chronicles, Epic’s operational database, running on InterSystems IRIS (successor to Caché) on Linux or AIX servers. Production runs on one active database server. IRIS mirroring replicates its data asynchronously and in near real time to a Report mirror, which serves operational reporting and the daily Clarity extract, and to a Disaster Recovery mirror at a second site. Depending on the design, a failover partner also protects the production server. During a short outage, users can keep working read-only on the Report mirror; for longer outages, the Disaster Recovery mirror takes over read/write.

Expertise needed

ODBA: IRIS administration, journaling and mirroring
Linux or AIX system administration and tuning
Storage performance and capacity planning
Failover, read-only and DR activation exercises

Lead rolesODBA

What runs here

Chronicles
InterSystems IRIS (successor to Caché)
Linux or AIX servers
Mirroring & failover
Report mirror
Disaster Recovery mirror
All-flash enterprise storage

When it fails

This is the patient record itself. Database latency is felt in every click; an outage moves the whole hospital into downtime procedures.

Analytics infrastructure

Epic reporting works on two levels. Operational reports such as Reporting Workbench work with near-real-time Chronicles data on the Report mirror; Radar dashboards combine such reports with other sources. For deeper analysis, a daily extract copies data from the Report mirror into Clarity, a relational database, and on into Caboodle, the data warehouse. Cogito is Epic’s umbrella for these tools. Epic is phasing out Oracle as a database for Cogito; new environments run on Microsoft SQL Server, on premises or in the cloud.

Expertise needed

Microsoft SQL Server administration
Planning the move away from Oracle where it still runs
ETL monitoring and load window management
Capacity planning for data growth

Lead rolesCogito / Analytics Infrastructure

What runs here

Reporting Workbench & Radar
Clarity
Caboodle
Daily extract from the Report mirror
Microsoft SQL Server
Oracle (being phased out)
Load windows & schedules

When it fails

A failed or late Clarity load means missing reports, wrong figures in quality and finance reporting, and decisions based on yesterday’s data.

Foundation

Everything above runs on a shared foundation of hypervisors, storage, networks, firewalls, backup and data centers or cloud.

Expertise needed

Virtualization and hypervisor operations
Storage and backup architecture
Network design, segmentation and firewalls
Infrastructure monitoring and lifecycle management

Lead rolesPlatform & Infrastructure, Network & Security

What runs here

Hypervisors
Enterprise storage
Data center network
Clinical Wi-Fi
Firewalls
Backup & recovery
Load balancing
Data centers & cloud
Monitoring

When it fails

A foundation problem rarely looks like one. It shows up as slow Epic, broken sessions or a failing interface, in several layers at once.

Across all layers

Four disciplines that touch every layer

Business continuity

Clinical work cannot stop when Epic is unavailable. Business Continuity Access (BCA) workstations and printed or locally stored downtime reports keep care going during planned and unplanned downtime. For short outages, users switch to a read-only copy on the Report mirror; for longer outages, the Disaster Recovery mirror is activated and takes over read/write. BCA workstations, their printers, their schedules and the switch to a mirror need to be tested regularly, just like the main system.

Security

Hardening, patching, segmentation, certificates and privileged access cover every layer. In a hospital, a security change can block a clinical workflow, so it has to be tested on representative workstations and rolled out in stages.

Monitoring and operations

Performance problems rarely sit where they first appear. Epic’s System Pulse shows live performance and system health and raises alerts, while System Pulse Analytics looks at long-term trends and capacity. Teams combine both with Citrix, network and infrastructure monitoring, shared measurement points from endpoint to database, clear on-call responsibilities and runbooks that the service desk can use.

Change and upgrades

Epic upgrades, Windows and Citrix updates, firmware and database maintenance all have to fit into tight maintenance windows. Good change management coordinates all teams around one timeline.

Germany · Austria · Switzerland

What is different in DACH

Epic was built in the United States. In the DACH region it has to connect to national eHealth infrastructures and run under European and Swiss regulation. Much of that work lands on the infrastructure teams.

Germany: Telematikinfrastruktur

Hospitals connect to the TI through TI connectors or the newer TI gateway. Health professional cards (eHBA) sit in card terminals at the workplace, which makes them an endpoint topic. Institution cards (SMC-B) usually sit in central card terminals; with the TI gateway, an HSM-B, a hardware security module in the gateway provider’s data center, can take their place. KIM messaging, the electronic patient record (ePA) and the e-prescription run on top. How each piece connects to Epic has to be designed and tested locally.

Switzerland: EPD and existing systems

Hospitals connect to the electronic patient record (EPD) through certified communities; the federal government plans to replace the EPD with the electronic health dossier (E-GD), which can go live in 2030 at the earliest. Billing often stays in an existing system next to Epic, such as SAP IS-H or a successor product, which adds interfaces and shared maintenance windows. Epic lists Swiss-specific integrations such as eMediplan and the exchange of case data with registration and billing systems. Data protection follows the revised Federal Act on Data Protection and cantonal law.

Austria: ELGA and e-card

ELGA, the national electronic health record, and the e-card system connect hospitals to national eHealth services. To our knowledge, no Austrian hospital has chosen Epic yet, so experience comes from projects in neighboring countries.

Security regulation

Germany has applied the NIS2 rules since December 2025, and Austria’s NISG 2026 obligations have applied since October 2026. Since April 2025, critical infrastructure in Switzerland, including listed hospitals, must report qualifying cyberattacks to the National Cyber Security Centre (NCSC, in German BACS) within 24 hours of discovery. Much of the evidence comes from infrastructure teams: logging, patching, segmentation and tested recovery.

Data residency and operations

Where data is stored, who operates the systems and who can access them from where are central questions in every DACH Epic project, whether the platform runs on premises, hosted or in the cloud.

Two working languages

Clinical users work with Epic in German or French, while technical documentation, vendor contacts and much of the expert exchange happen in English. Infrastructure teams move between both every day.

As of October 2026. Laws and national eHealth specifications change; check the current legal texts and specifications for your organization.

Technology fields