Overview

As modern data platforms evolve toward cloud-native and high-concurrency architectures, performance evaluation must go beyond single-query latency.

Today, enterprise databases are measured by their ability to:

• Scale under mixed OLAP and OLTP workloads
• Execute efficiently in distributed environments
• Maintain throughput stability under high concurrency

This benchmark provides a controlled, enterprise-grade comparison of SynxDB vs Greenplum (GP6 and GP7) using industry-standard TPC workloads.

---

Benchmark Scope

To simulate real-world production scenarios, three complementary benchmarks were selected:

• TPC-B – high-throughput transactional stress (TPS-focused)
• TPC-C – real-world OLTP workload simulation (tpmC)
• TPC-DS (1TB) – complex analytical queries (OLAP)

The objective is to evaluate:

• Distributed query execution efficiency
• Concurrency scalability limits
• System-level throughput stability

---

Test Environment (Hardware & Software)

To ensure fairness and reproducibility, all tests were conducted under identical infrastructure and cluster configurations.

Infrastructure Configuration

• Cloud Provider: AWS
• Instance Type: r5.4xlarge × 5 nodes

Total Resources

• 80 vCPU
• 640 GB Memory
• 20 TB GP3 Storage
• 50 Gbit/s Network

Cluster Topology

Both systems adopt a Massively Parallel Processing (MPP) architecture:

• 8 primary + 8 mirror segments per node
• Identical data distribution strategy
• Consistent shared-nothing MPP topology with identical segment-to-storage mapping

Software Versions

• Greenplum 6.27.1
• Greenplum 7.1.0
• SynxDB MPP

---

Workload Methodology

Industry-standard benchmarks were used to reflect realistic enterprise workloads.

TPC-B (High-Concurrency Stress)

• Focus: TPS (Transactions Per Second)
• Dataset: 10,000× standard scale
• Scenario: High-frequency read/write contention

TPC-C (OLTP Simulation)

• Focus: tpmC (new-order transactions per minute)
• Scale: 500 Warehouses
• Scenario: Order processing, inventory, and ACID transactions

TPC-DS (1TB OLAP)

• 99 complex SQL queries
• Includes joins, aggregations, subqueries, window functions
• Storage: AOCS + Zstd compression
• Data load: identical across systems (25 min @ 700 MB/s)

---

Benchmark Results

TPC-B Results: Transaction Throughput (TPS)

Concurrency	SynxDB	GP6	GP7
1	100.5	102.1	109.4
5	544.2	525.8	550.9
10	1074.3	988.3	1118.2
15	1519.6	1414	1484.3
20	2132.8	1745.8	2132.3
30	2542.9	2342	2460
40	2472.5	2460	2437.2

Analysis

At lower concurrency levels, all systems perform similarly. However, as concurrency increases:

• SynxDB demonstrates stronger scaling consistency
• Peak performance occurs at 30 concurrency (2542.9 TPS)
• Greenplum shows efficiency degradation under load

👉 This indicates better contention handling and resource scheduling in SynxDB.

---

TPC-C Results: OLTP Throughput (tpmC)

Concurrency	SynxDB	GP6	GP7
1	1186	278	399
2	2284	—	454
5	5331	—	452
10	9239	—	453

Analysis

• GP6 fails beyond concurrency = 1 due to transaction timeouts
• GP7 plateaus at ~450 tpmC regardless of scaling
• SynxDB achieves near-linear growth

👉 At concurrency 10, SynxDB delivers ~20× higher throughput than GP7

This highlights a fundamental architectural limitation in Greenplum under OLTP pressure.

---

TPC-DS Results: Analytical Query Performance (1TB)

Scenario	SynxDB	GP6	GP7
Single Query Stream	5335s	6834s	6088s
5 Concurrent Queries	21125s	28255s	24750s

Analysis

• 21.9% faster than GP6 (single)
• 12.3% faster than GP7 (single)
• 25.2% faster than GP6 (concurrent)
• 14.6% faster than GP7 (concurrent)

Performance gains are primarily driven by:

• Reduced inter-node data shuffle
• More efficient distributed execution planning
• Better pipeline execution

👉 In large-scale OLAP workloads, network cost dominates performance, and SynxDB handles it more efficiently.

---

Final Conclusion

Across all three benchmarks—TPC-B, TPC-C, and TPC-DS—the results consistently demonstrate that:

SynxDB delivers superior performance, scalability, and stability compared to Greenplum at scale.

---

Key Findings

• Up to 20× higher OLTP throughput (TPC-C)
• Up to 25.2% faster analytical performance (TPC-DS)
• More stable scaling under increasing concurrency
• No observable performance ceiling in tested scenarios

---

What This Means for Enterprises

Greenplum shows clear limitations in modern workloads:

• Throughput stagnation under concurrency
• Transaction instability under stress
• Inefficient distributed execution at scale

SynxDB, by contrast, provides:

• Linear scalability
• Stable mixed-workload execution
• Production-grade efficiency and architectural flexibility

Related Resources

• What the Licensing Change Means for Open Source Users
• Why Apache Cloudberry Is the Most Natural Open Source Alternative to Greenplum