Compile policy into a local real-time representation
PostgreSQL remains authoritative for administration, while a C++ loader/compiler transforms the configured rules into an optimized in-memory form. Per-call checks never require a PostgreSQL query.
Networking and High-Performance Systems
SentinelVoice - High-Rate Telecom Fraud-Prevention Filtering Gateway
A real-time mobile-network filtering gateway with custom C++ in-memory matching, SIGTRAN/CAMEL integration, more than 10 million call checks per second against 10 million policy entries, and multi-operator production deployment.
CAMEL and SIGTRANC++ systemsIn-memory matchingTelecom policy enforcement
The system
SentinelVoice connects to a mobile-network call-control environment through SIGTRAN and CAMEL integration. For each relevant call attempt, it extracts the calling and called numbers, evaluates them against customer-maintained policy records, and returns one of three enforcement actions: allow the call, drop it, or forward it to an IVR.
The gateway does not determine whether a call is fraudulent. Customer teams and surrounding systems create the blacklist, policy, and pattern records. OPTIME engineered the deterministic, high-performance real-time enforcement layer that applies those supplied decisions inside the signaling path.
Policy records can contain exact values and pattern-based rules. At the required rates, scanning a flat list or evaluating a conventional collection of regular expressions for every call would have been computationally unsuitable, so the real-time design required a different local representation and matching approach.
Engineering relationship
Across a two-year specialized telecom program, OPTIME engineered the signaling integration, the higher protocol/application functionality needed by the gateway, the C++ rule-loading and matching engine, the administration plane, production integration, and performance validation.
OPTIME used the Dialogic SIGTRAN stack as the lower transport/signaling foundation and implemented the application-specific functionality required above MTP3 for the CAMEL-integrated gateway. The public architecture intentionally stops short of proprietary protocol flows and customer MSC configuration.
Engineering constraints
- Meet an original requirement of approximately one million call checks per second against roughly 100,000 policy records.
- Support exact-value and pattern-aware rules without sequentially scanning every rule for every call.
- Keep PostgreSQL out of the per-call real-time path while retaining it as the canonical policy store.
- Integrate with the Dialogic SIGTRAN foundation and the MSC/CAMEL call-control environment.
- Return allow, drop, or forward-to-IVR actions with very low processing overhead.
- Remain practical to deploy in ordinary virtualized Linux server environments without specialized accelerators.
- Protect customer signaling, policy, matching, and deployment details while demonstrating measured performance.
What OPTIME engineered
- Dialogic SIGTRAN integration and the higher signaling/application functionality required above MTP3 for the gateway.
- CAMEL/MSC call-control integration and calling/called-number extraction at a generalized public boundary.
- A C++ rule loader/compiler that reads configured policies from the canonical PostgreSQL store.
- A custom optimized local in-memory representation for exact and pattern-aware rule evaluation.
- A proprietary C++ matching algorithm designed for multi-million-check-per-second execution.
- Immediate allow, drop, and forward-to-IVR enforcement actions returned to the surrounding call-control path.
- A Node.js/TypeScript administration backend and React management console.
- Rule-management synchronization from PostgreSQL into the compiled in-memory real-time representation.
- Linux and virtual-machine deployment integration plus performance and production validation.
Architecture
Real-time plane - MSC / CAMEL
Relevant mobile-network call attempts enter the filtering workflow through the customer’s CAMEL-integrated MSC environment.
Real-time plane - SIGTRAN foundation
The Dialogic SIGTRAN stack provides the lower signaling foundation used by the gateway.
Real-time plane - OPTIME protocol/application layer
OPTIME-built functionality above MTP3 extracts the calling and called numbers required for policy enforcement.
Real-time plane - in-memory matching
The C++ engine evaluates exact and pattern-aware policy entries using a proprietary optimized local representation.
Real-time plane - enforcement action
The gateway returns allow, drop, or forward-to-IVR without performing independent fraud analysis.
Management plane - React administration
Operators manage filtering and policy records through the administration console.
Management plane - Node.js / TypeScript
Backend services validate and coordinate policy administration outside the per-call processing path.
Management plane - PostgreSQL
PostgreSQL remains the canonical administrative and policy store rather than a database queried for every call.
Rule loading and compilation
The C++ loader transforms configured records into the optimized in-memory representation used by the real-time plane.
Key engineering decisions
Avoid naive pattern scanning
The rule set includes exact values and patterns. A conventional sequential or regex-style scan across the complete set would not meet the required rate, so OPTIME designed a proprietary pattern-aware matching architecture.
Separate policy creation from enforcement
Customer and surrounding systems decide which policy records represent fraud-prevention actions. SentinelVoice applies those supplied records deterministically and does not infer fraud.
Remain deployable in virtualized infrastructure
The achieved throughput did not depend on specialized acceleration hardware, allowing the production gateway to run in virtualized Linux server environments.
Performance engineering
The original requirement was approximately one million call checks per second against roughly one hundred thousand policy records. OPTIME’s optimized in-memory matching architecture ultimately exceeded ten million checks per second against a ten-million-entry rule set.
The design keeps administration and canonical policy storage out of the signaling hot path. Rule compilation happens when policy changes are loaded; call processing uses only the prepared local representation and immediate enforcement logic.
Verified capability
The delivered gateway continues operating in production across multiple mobile operators. It enforces customer-supplied filtering policy through CAMEL/SIGTRAN integration and runs in ordinary virtualized server environments rather than requiring specialized accelerator hardware.
Verified result
OPTIME delivered a production filtering gateway that exceeded ten million call checks per second against ten million policy entries while preserving a strict boundary between customer fraud-policy decisions and deterministic real-time enforcement.
Verified metrics
≈1M checks/s · ≈100K records
The customer’s starting performance and policy-set requirement.
>10M checks/s · 10M entries
Validated matching throughput and rule-set size for the optimized architecture.
Technology & Engineering Role
- C++
- Real-time signaling integration, rule loading, optimized matching, and call-enforcement decisions.
- Custom matching algorithm
- Pattern-aware policy evaluation at multi-million-check-per-second rates.
- Custom in-memory representation
- Avoids database queries and naive sequential pattern evaluation in the real-time path.
- Dialogic SIGTRAN
- Lower telecom transport and signaling foundation.
- CAMEL
- Mobile-network call-control environment for the filtering gateway.
- PostgreSQL
- Canonical administrative policy and filtering-rule store.
- Node.js / TypeScript
- Administration and rule-management backend.
- React
- Administration console for policy management.
- Virtualized Linux environment
- Production deployment without specialized accelerator hardware.
Related engineering
Telecom & Real-Time Communications EngineeringAccelerated Computing & Performance EngineeringNetworking, SDN & Open Infrastructure EngineeringHigh-Performance Packet and Protocol ProcessingOpen Mobile Core and Private 4G/5G NetworksNetworking and InfrastructureAccelerated ComputingSecurity and Identity
~500 GbpsMobileScope - Mobile-Core Traffic and GTP Analytics
>1 TbpsPacketLens - High-Throughput IP Traffic Intelligence Platform
~500 GbpsMobileScope - Mobile-Core Traffic and GTP Analytics>1 TbpsPacketLens - High-Throughput IP Traffic Intelligence Platform