Real-Time Event Streaming with IBM Confluent Platform for Z and LinuxONE: Performance Validation at Enterprise Scale - ILOG SA
Financial Performance:
- No specific financial metrics (revenue, earnings, profit margins, debt, cash flow) were disclosed as this is a technical product webinar rather than a corporate earnings call.
- The session highlighted potential cost savings for customers through reduced egress fees, fewer physical cores required, and consolidation of Kafka brokers, leading to lower infrastructure management overhead.
Guidance and Future Outlook:
- Roadmap: Native ZOS connectors and ZOS brokers are planned for future release.
- Timeline: Business Event Refinery architecture (Layer 4) is scheduled for preview in Q1; native connectivity for Z subsystems (DB2, zIIP, IMS, MQ, zDIIH) became generally available (GA) in November.
- Strategic Goal: To position IBM Z as a first-class producer/consumer of real-time events to support the rapid adoption of agentic AI by 2028.
- Next Steps: Encouraging customers to engage in discovery workshops, Proof of Concepts (POCs), and customer-specific performance benchmarks.
Business Segments and Product Lines:
- Product: IBM Confluent Platform for Z and Linux on Z.
- Key Architecture: "Business Event Refinery" which transforms raw fragmented signals from Z subsystems into trusted business events.
- Connectivity: Native support for z/OS subsystems including CICS (referred to as Kix/KICS), DB2, IMS, MQ, zDIIH, and SMF logs.
- Capabilities: Transactional consistency at commit, zero COBOL application rewrites, schema validation (Avro from COBOL copybooks), lineage tracking, and hardware encryption.
- Deployment Options: Runs on ZCX, Linux on Z, OpenShift, and supports containerized virtualized deployments (OCP virtualization) starting in the near future.
Market and Competitive Landscape:
- Market Trend: Shift toward agentic AI requires real-time, trusted data; currently, 91% of organizations do not fully understand AI dependencies or data lineage.
- Use Cases: Real-time fraud detection (blocking transactions within milliseconds), live customer views, next-best-action personalization, trade finance (sub-millisecond latency), and hybrid cloud disaster recovery (DR/HA).
- Competitive Position: IBM Z positions itself as more efficient than x86/public cloud for high-volume, low-latency event streaming due to vertical scaling capabilities and hardware cryptography.
- Integration: Supports hybrid broker architectures to link Z-based clusters with cloud or x86 platforms via Confluent Cluster Linking.
Risks and Challenges:
- Data Latency: Traditional batch processing results in "stale" data, causing missed opportunities for fraud prevention or real-time decisioning.
- Complexity: Fragmented point-to-point integrations lead to governance issues, lack of data lineage, and high maintenance overhead.
- Context Loss: Raw row-level data streaming lacks business context (intent, triggers), requiring ad hoc downstream processing.
- Platform Limits: Performance on Linux on Z is sensitive to CPU over-commitment and requires specific parameter tuning compared to x86.
Management Commentary and Tone:
- Tone: Confident, technical, and solution-oriented; emphasized that AI agents require real-time, governed data which Z provides.
- Key Message: IBM Z is moving from a passive record-keeping system to an active, real-time event producer.
- Confidence: Management expressed strong confidence in the performance benchmarks, highlighting the ability to achieve similar results with fewer brokers and cores compared to x86.
Analyst Questions and Answers:
- Question: What are common use cases beyond fraud?
- Answer: Trade finance (sub-millisecond latency), operational data stores (MongoDB/PostgreSQL on Z), hybrid cloud eventing via cluster linking, and DR/HA scenarios where Z serves as the primary or failover site.
- Question: How does the platform handle data from different Z subsystems?
- Answer: Native connectors capture signals from CICS, IMS, DB2, etc., which are then correlated and enriched within the "Business Event Refinery" on Z before being published as events.
- Question: How will performance scaling be tested?
- Answer: Upcoming tests will utilize 100Gb integrated network adapters (Z17 Linux 1.5) and containerized Confluent Kafka deployments.
Other Key Points:
- Benchmark Results: Linux on Z demonstrated 6x to 8x lower P99 latency compared to open-source Kafka at 1GB/s throughput; required 30% fewer brokers and 6x to 8x fewer processor cores for equivalent workloads at high scale.
- Performance Nuance: Increasing vCPUs beyond a certain threshold on Linux on Z can decrease performance due to platform vertical-scaling nature; parameter tuning (e.g., linger time) differs from x86.
- Security: Leverages IBM Z hardware cryptography for in-flight and at-rest data security without performance trade-offs.
- No Application Rewrites: The architecture allows for event streaming without modifying existing COBOL applications.
- Company
- ILOG SA
- Keywords
- ILOG SA