Cloud Native API Gateway supports purchasing based on node specifications and quantity, and also supports adjusting node specifications and quantity after purchase. This document introduces the performance metrics of Cloud Native API Gateway Kong. Select an appropriate resource size based on your business requirements.
Gateway Operating Thresholds
The gateway operating threshold is described as follows:
Recommended threshold: The CPU utilization is 30%, and the memory utilization is 30%.
Warning threshold: The CPU utilization is 60%, and the memory utilization is 60%.
In routine operations, it is recommended that the resource utilization be 30% to 60%. This maintains high resource utilization and can cope with sudden request surges. Under the warning threshold, the gateway can achieve full Service Level Agreement (SLA) assurance. When the operating threshold exceeds the warning threshold, the gateway enters a heavy load status, and the request success rate and latency may be affected. It is recommended that alarms be configured.
Gateway Capacity Metrics
The node specification and quantity determine the gateway capacity. Increasing the node specification or adding nodes can improve the gateway capacity. If the node specification remains unchanged, the gateway capacity maintains linear growth as the node quantity increases. Gateway capacity metrics are described as follows:
Number of connections: maximum number of connections supported by the gateway.
Number of new HTTP connections per second: maximum number of non-persistent HTTP connections the gateway can create per second.
Number of new HTTPS connections per second: maximum number of non-persistent HTTPS connections the gateway can create per second.
Node specification |
| 2-core and 4 GB | 4-core and 8 GB | 8-core and 16 GB | 16-core and 32 GB |
Node Quantity |
| 1 | 1 | 1 | 1 |
CPU utilization | Recommended threshold | 30% | 30% | 30% | 30% |
| Warning threshold | 60% | 60% | 60% | 60% |
Memory Utilization | Recommended threshold | 30% | 30% | 30% | 30% |
| Warning threshold | 60% | 60% | 60% | 60% |
Number of connections | Recommended threshold | 24000 | 48000 | 96000 | 192000 |
| Warning threshold | 48000 | 96000 | 192000 | 384000 |
Number of New HTTP Connections Per Second | Recommended threshold | 2000 | 4000 | 8000 | 16000 |
| Warning threshold | 4000 | 8000 | 16000 | 32000 |
Number of New HTTPS Connections Per Second | Recommended threshold | 600 | 1200 | 2400 | 4800 |
| Warning threshold | 1200 | 2400 | 4800 | 9600 |
Gateway Performance Data
Test scenario: The client sends a request to the gateway, which forwards it to the backend server, and the backend server returns a response directly.
Test conditions: different types of requests, connections, and response sizes.
Test result: The following table describes the client request QPS, and the average request latency is less than 5 ms.
CPU Utilization and Memory Utilization: Within Recommended Threshold (30%) |
|
|
|
|
|
|
|
|
|
|
Protocol Type | Connection Type | Response Size | 2-core and 4 GB |
| 4-core and 8 GB |
| 8-core and 16 GB |
| 16-core and 32 GB |
|
|
|
| 3-node | 5-node | 3-node | 5-node | 3-node | 5-node | 3-node | 5-node |
HTTP | Non-persistent Connection | 512B | 6500 | 10800 | 13000 | 21600 | 25000 | 41650 | 50000 | 83300 |
|
| 1KB | 5800 | 9600 | 11600 | 19200 | 23200 | 38650 | 46400 | 77300 |
|
| 10KB | 4500 | 7500 | 9000 | 15000 | 20000 | 33300 | 40000 | 66650 |
| Persistent connection | 512B | 7000 | 11600 | 14000 | 23200 | 28800 | 48000 | 57600 | 96000 |
|
| 1KB | 6700 | 11100 | 13400 | 22200 | 27000 | 45000 | 54000 | 90000 |
|
| 10KB | 6000 | 10000 | 12000 | 20000 | 24000 | 40000 | 48000 | 80000 |
HTTPS | Non-persistent Connection | 512B | 1800 | 3000 | 3600 | 6000 | 7800 | 13000 | 15600 | 26000 |
|
| 1KB | 1700 | 2800 | 3400 | 5600 | 7500 | 12500 | 15000 | 25000 |
|
| 10KB | 1600 | 2600 | 3200 | 5200 | 7100 | 11800 | 14200 | 23650 |
| Persistent connection | 512B | 6450 | 10750 | 12900 | 21500 | 26000 | 43300 | 52000 | 86650 |
|
| 1KB | 6200 | 10300 | 12400 | 20650 | 25200 | 42000 | 50400 | 84000 |
|
| 10KB | 5500 | 9150 | 11000 | 18300 | 22000 | 36650 | 44000 | 73300 |