X Platform 3.16.34

Performance results for X Platform 3.16.34, based on testing conducted in February 2025.

Test Configuration

• X Platform Version: 3.16.34

• Java Runtime: Oracle Java 8

• Message Encoding: Xbuf2

• Cluster Configuration: Primary + Backup with persistence and replication

• Test Hardware: Intel Xeon Gold 6334 (8-Core, 3.6 GHz), 128GB RAM, InfiniBand network

• Message Rate (Latency Tests): 10,000 messages/second

• Message Rate (Throughput Tests): As fast as possible (saturated)

See the Test Description for complete test methodology and hardware specifications.

Latency Results

All latency numbers are in microseconds (µs). Round-trip wire latency (~23µs on unoptimized network) is included in all results.

Indirect Message Access | Latency Optimization

Configuration: Xbuf2.Indirect | OptimizeFor=Latency | Message Rate=10,000/sec

Message data accessed via POJO setter/getter methods.

CPU Config	# CPUs	50th % (µs)	99th % (µs)	99.9th % (µs)
MinCPU	1	40.90	43.31	44.95
Default	4	30.55	34.85	38.27
MaxCPU	6	32.44	36.61	40.23

Best Configuration: Default (4 CPUs) - 30.55µs median latency

Direct Message Access | Latency Optimization

Configuration: Xbuf2.Direct | OptimizeFor=Latency | Message Rate=10,000/sec

Message data accessed via serializer/deserializer objects (zero-copy access).

CPU Config	# CPUs	50th % (µs)	99th % (µs)	99.9th % (µs)
MinCPU	1	37.72	40.07	41.43
Default	4	27.34	30.56	34.04
MaxCPU	6	29.09	33.08	47.14

Best Configuration: Default (4 CPUs) - 27.34µs median latency

Key Insight: Direct message access provides ~10% lower latency than indirect access by eliminating POJO method call overhead.

Throughput Results

Throughput measured in messages per second. Test mode: saturated load (as fast as possible).

Indirect Message Access | Throughput Optimization

Configuration: Xbuf2.Indirect | OptimizeFor=Throughput | Message Rate=[As Fast As Possible]

CPU Config	# CPUs	Throughput (msgs/sec)
MinCPU	1	117,124
Default	4	90,974
MaxCPU	6	56,419

Best Configuration: MinCPU (1 CPU) - 117,124 msgs/sec

Direct Message Access | Throughput Optimization

Configuration: Xbuf2.Direct | OptimizeFor=Throughput | Message Rate=[As Fast As Possible]

CPU Config	# CPUs	Throughput (msgs/sec)
MinCPU	1	282,038
Default	4	281,758
MaxCPU	6	107,336

Best Configuration: MinCPU (1 CPU) - 282,038 msgs/sec

Key Insight: Direct message access provides 2.4x higher throughput than indirect access by eliminating serialization overhead.

Performance Analysis

Latency Characteristics

1. Optimal CPU Configuration: 4 CPUs (Default) provides best latency across both access methods

   • Balances parallelization benefits against thread coordination overhead

   • 25-30% better than MinCPU configuration

2. Direct vs Indirect Access: Direct access reduces latency by ~10%

   • Median: 27.34µs (Direct) vs 30.55µs (Indirect)

   • Avoids POJO method call overhead and intermediate object creation

3. Tail Latency: 99.9th percentile latencies remain within 1.5x of median

   • Indicates consistent, predictable performance

   • Good mechanical sympathy with modern hardware

Throughput Characteristics

1. Optimal CPU Configuration: MinCPU (1-2 CPUs) provides best throughput

   • Lightweight message handler benefits from single-threaded execution

   • Thread handoff overhead exceeds parallelization benefits in this test

   • Note: Real applications with heavier business logic may benefit from more CPUs

2. Direct vs Indirect Access: Direct access provides 2.4x throughput improvement

   • MinCPU: 282K msgs/sec (Direct) vs 117K msgs/sec (Indirect)

   • Zero-copy access eliminates serialization bottleneck

3. CPU Scaling: Throughput decreases with more CPUs in this lightweight test

   • High-performance disk and zero-cost outbound messaging (in-process driver)

   • Thread coordination overhead exceeds benefits

   • Applications with heavier processing loads will see different scaling characteristics

Tuning Recommendations

For Lowest Latency

1. Use Direct message access (serializer/deserializer objects)

2. Configure 4 CPUs (Default configuration)

3. Enable latency optimization mode

4. Expected: ~27µs median, ~31µs 99th percentile

For Highest Throughput

1. Use Direct message access (serializer/deserializer objects)

2. Configure MinCPU (minimal thread count)

3. Enable throughput optimization mode

4. Expected: ~280K msgs/sec

Application-Specific Considerations

• Heavier business logic: May benefit from more CPUs (Default or MaxCPU)

• Complex message transformations: Direct access provides larger benefits

• Network-limited scenarios: VMA enablement can further reduce latency

• Multiple microservices per host: CPU isolation becomes critical

Next Steps

• Review Test Description for complete test methodology

• Return to Performance Overview

• See Threading Configuration for CPU tuning

• Learn about Direct Serialization for optimal performance

• Explore Thread Affinitization for CPU pinning