1. What Is 3GPP EPC Control and User Plane Separation Architecture (CUPS, TS 23.214)?
The 3GPP EPC Control and User Plane Separation Architecture separates mobile gateway functions that manage sessions from functions that process subscriber traffic. CUPS means Control and User Plane Separation; EPC means Evolved Packet Core; and 3GPP is the Third Generation Partnership Project. TS 23.214 is the architectural Technical Specification. This is an engineering reference architecture, not a management scoring model.
For telecommunications operators, the executive question is whether separating gateway control from packet processing justifies a different capacity, location, and investment model. Consultants should use CUPS to structure those choices—not assume that adopting the architecture automatically creates a business case.
2. Origin and Background
Origin: CUPS was standardized by 3GPP, with architectural responsibility in its System Architecture working group SA2, for Release 14. The preceding study was TR 23.714; TS 23.214 was created in June 2016 and placed under change control in September 2016. It was not created by an individual management theorist or consulting firm.
The objective was flexible placement and scaling of EPC gateway functions while retaining their overall functionality. Its original context was EPC evolution, rather than replacement of the EPC with a standalone 5G core.
3. How 3GPP EPC Control and User Plane Separation Architecture (CUPS, TS 23.214) Works
The acronym describes an architectural separation, not four implementation stages:
- C — Control: Gateway decisions concerning sessions, bearers, and packet-treatment rules.
- U — User: Subscriber traffic carrying application data.
- P — Plane: The logical distinction between signaling and traffic processing.
- S — Separation: Distinct functions connected through standardized interfaces.
The architecture separates three types of EPC function. The suffix -C identifies the control-plane function; -U identifies the user-plane function.
| Original function | Separated functions | Control-to-user interface |
|---|---|---|
| Serving Gateway, or SGW | SGW-C and SGW-U | Sxa |
| Packet Data Network Gateway, or PGW | PGW-C and PGW-U | Sxb |
| Traffic Detection Function, or TDF, where deployed | TDF-C and TDF-U | Sxc |
Combined SGW/PGW deployments remain possible. CUPS does not replace the Mobility Management Entity or separate every EPC component.
The Sx interfaces use the Packet Forwarding Control Protocol, or PFCP, specified in TS 29.244. Control functions establish, modify, and delete user-plane sessions. Rules govern packet detection, forwarding, quality-of-service enforcement, usage reporting, and buffering where needed. The user plane therefore retains session-specific rules; separation does not make it stateless.
For deployment planning, control functions can remain centralized while user-plane functions are distributed. The practical implication is that local traffic delivery requires examining the entire route: moving SGW-U closer to subscribers is insufficient if PGW-U and the application destination remain distant.
4. When to Use 3GPP EPC Control and User Plane Separation Architecture (CUPS, TS 23.214)
CUPS is especially relevant when an operator needs additional packet-processing capacity without proportionate expansion of gateway control functions, or wants different user-plane locations for different services. Its ability to support flexible placement and separate scaling is the architectural advantage.
Useful decision questions include:
- Should broadband traffic use regional gateway user planes rather than centralized processing?
- Which services genuinely benefit from regional packet-data connectivity?
- Does further EPC investment make sense alongside the planned 5G-core migration?
Start with busy-hour traffic, packet rates, concurrent sessions, session transaction rates, routing dependencies, vendor capabilities, and site costs. Bring together core engineering, transport, charging, security, network operations, and finance.
It is a poor fit for a reseller that does not control gateway architecture, or for a problem dominated by radio congestion. It can also mislead when the analysis assumes traffic can exit locally despite centralized service dependencies.
Separate the architecture workshop from evidence gathering, integration testing, and production migration. Follow-on program management support should coordinate vendor deliverables, transport changes, charging integration, and cutover dependencies.
5. How to Apply 3GPP EPC Control and User Plane Separation Architecture (CUPS, TS 23.214): Step-by-Step
- Clarify the decision and scope.
Specify whether the objective is capacity expansion, regional traffic delivery, gateway renewal, or migration preparation. Define the investment horizon, expected EPC lifetime, geographic boundaries, services, and gateway functions included.
- Gather demand and dependency data.
Collect busy-hour throughput, packet-size distributions, concurrent sessions, bearer activity, and session creation or modification rates. Map charging, policy, roaming, security, and application-routing dependencies. Reconcile network telemetry with vendor capacity definitions.
- Define the units of analysis.
Assess each region, gateway role, and service or access point name, commonly abbreviated APN. Distinguish broadband, enterprise connectivity, and voice-related traffic where their routing or operational requirements differ.
- Construct the architecture artifact.
Draw control-plane pools, candidate user-plane sites, Sx connections, subscriber traffic paths, and external network connections. Mark failure domains and deployment boundaries. Accompany the diagram with a capacity table and dependency register.
- Compare deployment alternatives.
Evaluate centralized, selectively distributed, and more extensively distributed options. Calculate capacity under failure conditions, not just normal operation. Include transport, licensing, infrastructure, redundancy, integration, and ongoing support in the cost comparison.
- Translate the preferred option into actions.
Define vendor requirements, infrastructure changes, pilot scope, migration waves, and acceptance gates. Include session establishment, packet treatment, charging reconciliation, service continuity, and rollback testing rather than approving on throughput alone.
- Test sensitivities.
Vary traffic growth, packet rates, session churn, site failures, transport availability, and EPC retirement timing. Identify conditions under which regional deployment stops being attractive. Avoid presenting a single forecast as certainty.
- Align stakeholders and iterate.
Have engineering, operations, finance, security, charging, and vendors challenge the assumptions. Assign decision owners and operational acceptance criteria. Update the design after laboratory evidence and pilot results.
The resulting design should feed into systems implementation oversight, with explicit responsibility for integration evidence, acceptance testing, migration readiness, and rollback authority.
6. Example: 3GPP EPC Control and User Plane Separation Architecture (CUPS, TS 23.214) in Action
Illustrative case: North Harbor Mobile is a fictional operator serving three regions through two centralized gateway sites. Broadband throughput is growing faster than gateway session transactions. Its CTO wants additional capacity without committing immediately to a complete core replacement.
The team selects CUPS to compare retaining centralized gateways with introducing regional user-plane processing. Its units of analysis are region, APN, and gateway role. Inputs include busy-hour measurements, packet rates, session activity, routing maps, vendor capabilities, and fully loaded site costs.
The analysis shows that one region has sufficient broadband demand and existing regional peering to justify local packet-data connectivity. Enterprise traffic still depends on centralized security services. A uniform rollout would therefore add sites without equivalent benefits everywhere.
North Harbor retains gateway control functions at its central sites and pilots combined SGW/PGW user-plane functions in the selected region, connected through PFCP. Eligible new broadband sessions use the pilot deployment; existing sessions are handled through an explicitly tested migration plan.
Operations and charging teams validate bearer procedures, packet treatment, usage reconciliation, and recovery behavior. Expansion remains conditional on the pilot’s acceptance gates and a business case tested against slower traffic growth. The decision is selective distribution, not “edge everywhere.”
7. Strengths and Limitations
Strengths
- Separates capacity decisions: Control and packet-processing resources can be expanded according to different demand patterns.
- Supports placement choices: User-plane deployment can reflect traffic geography and service requirements.
- Provides a common technical language: Standardized function boundaries help structure operator–vendor design discussions.
Limitations
CUPS does not determine the optimal number of sites, hardware configuration, licensing model, or migration schedule. Nor does architectural separation guarantee uninterrupted recovery. Implementations differ in their redundancy mechanisms, including capabilities beyond the EPC standard; established traffic and new-session handling require separate failure assessments.
For hosted deployments, assess cloud and data centers alongside the gateway design. Require clear ownership of host failures, transport reachability, observability, and service recovery before counting operational benefits.
Most importantly, evaluate the remaining useful life of the EPC investment. A technically attractive deployment may still be commercially unattractive if its benefits arrive after the planned transition.
8. Common Pitfalls and How to Avoid Them
- Using one capacity metric: Gigabits per second alone conceal packet-rate, session-state, and signaling demands. Build separate workload profiles and test realistic traffic mixes.
- Equating distribution with local delivery: A nearby serving gateway does not establish a nearby packet-data exit. Trace the complete path through PGW-U and dependent services.
- Assuming transparent session movement: Do not promise disruption-free migration from a diagram. Validate supported procedures, session treatment, maintenance windows, and rollback with the selected implementation.
- Treating standard interfaces as interoperability proof: Specify releases, supported features, rule behavior, and acceptance tests. Require demonstrated integration rather than relying on a compliance statement.
- Comparing against an unrealistic baseline: Include redundancy, idle regional capacity, licenses, transport, and operational workload. Compare against credible centralized expansion and 5G-transition alternatives.
9. How 3GPP EPC Control and User Plane Separation Architecture (CUPS, TS 23.214) Relates to Other Frameworks
Total cost of ownership and scenario analysis
Use total cost of ownership alongside CUPS to evaluate deployment economics. Start with service requirements and demand scenarios, construct feasible architectures, and then compare lifecycle costs. Prefer this combination when the question is where to invest—not merely whether separation is technically possible.
Failure Mode and Effects Analysis
Use Failure Mode and Effects Analysis after drawing the candidate architecture. Examine control-function failure, user-plane failure, transport partition, and inconsistent session state. Convert the findings into redundancy decisions, monitoring requirements, and recovery tests.
5G-core architecture
The 5G core carries forward control/user-plane separation through the Session Management Function and User Plane Function, but it introduces a different architecture. CUPS can support an EPC-to-5G migration path; it does not itself create a standalone 5G core. Compare the two before committing to long-lived gateway investments.
10. Key Takeaways
- CUPS is a standardized EPC architecture, not a strategy matrix.
- Use it to evaluate gateway scaling, placement, and investment choices.
- Analyze control demand and packet-processing demand separately.
- Trace complete service paths before claiming regional delivery benefits.
- Make migration, recovery, and lifecycle economics explicit approval gates.
11. FAQs About 3GPP EPC Control and User Plane Separation Architecture (CUPS, TS 23.214)
Is EPC CUPS the same as a 5G core?
No. EPC CUPS separates existing gateway functions. A 5G core uses different network functions and architectural arrangements, although it retains the separation principle.
Does CUPS automatically reduce latency?
No. Distributed user-plane placement creates an opportunity to shorten traffic paths. The benefit depends on gateway selection, packet-data connectivity, and the application destination—not proximity alone.
What is the difference between TS 23.214 and TS 29.244?
TS 23.214 defines the architectural separation and functional relationships. TS 29.244 specifies PFCP, the protocol used between control-plane and user-plane functions.
How long does a CUPS assessment take?
Scope the assessment separately from implementation. Timing depends on telemetry quality, topology complexity, vendor evidence, service dependencies, and available testing environments. A workshop can frame options; it cannot establish production readiness.
