Capacity and performance specifications
Capacity thresholds, QPS benchmarks, and quota limits for Cloud-native API Gateway, covering both Dedicated and Serverless instances.
This document covers capacity thresholds, QPS benchmarks, and quota limits for both Dedicated instances and Serverless instances of Cloud-native API Gateway.
Serverless instances are currently in public beta. Quotas listed for Serverless are beta-period values and may change upon commercial release.
Instance types
Instance type | Resource model | Capacity management | Use cases |
Dedicated instance | Dedicated physical resources per user | Choose a specification; higher spec = higher capacity | High-frequency stable workloads, compliance auditing, deep customization |
Serverless instance | Platform-managed multi-tenant shared pool | Unified platform capacity ceiling; no specification choice | Variable traffic, small-to-medium scale, quick onboarding |
Capacity thresholds (Dedicated instances)
The following table lists capacity thresholds per gateway specification. Full SLA coverage applies when metrics stay below the warning threshold. For critical workloads, keep metrics below the secure threshold.
Secure threshold: The gateway maintains high throughput and low latency even if traffic doubles.
Warning threshold: Exceeding this level may cause increased latency and stability risks during traffic surges.
The
apigw.dev.x1specification is a single-node deployment with no SLA guarantee, intended for testing only. You must use a multi-node specification for production.The SLA does not cover request failures when CPU or memory exceeds the warning threshold. The gateway provides alarms for these thresholds. Monitor gateway load and triggered alarms closely.
Gateway specification | Client connections | New HTTPS connections | CPU utilization | Memory usage | ||||
Secure threshold | Warning threshold | Secure threshold | Warning threshold | Secure threshold | Warning threshold | Secure threshold | Warning threshold | |
| 12,000 | 24,000 | 400 | 800 | 30% | 60% | 75% | 75% |
| 24,000 | 48,000 | 800 | 1,600 | 30% | 60% | 75% | 75% |
| 48,000 | 96,000 | 1,600 | 3,200 | 30% | 60% | 75% | 75% |
| 96,000 | 192,000 | 3,200 | 6,400 | 30% | 60% | 75% | 75% |
| 192,000 | 384,000 | 6,400 | 12,800 | 30% | 60% | 75% | 75% |
| 384,000 | 768,000 | 12,800 | 25,600 | 30% | 60% | 75% | 75% |
| 576,000 | 1,152,000 | 19,200 | 38,400 | 30% | 60% | 75% | 75% |
| 768,000 | 1,536,000 | 25,600 | 51,200 | 30% | 60% | 75% | 75% |
| 1,536,000 | 3,072,000 | 51,200 | 102,400 | 30% | 60% | 75% | 75% |
| 2,304,000 | 4,608,000 | 76,800 | 153,600 | 30% | 60% | 75% | 75% |
| 3,072,000 | 6,144,000 | 102,400 | 204,800 | 30% | 60% | 75% | 75% |
Serverless instances use platform-managed resources with no specification choice. See the "Serverless instance capacity thresholds" section for per-tenant thresholds.
QPS performance (Dedicated instances)
Gateway QPS throughput depends on factors such as response size, HTTPS usage, and gzip compression. The following tables list pessimistic QPS values at 30% CPU (secure threshold).
HTTPS new connections consume significant CPU resources. For burst HTTPS scenarios, refer to the short connection + HTTPS data to evaluate gateway capacity.
Short connections
Response size | HTTPS | gzip | dev.x1 | small.x1 | small.x2 | small.x4 | medium.x1 | medium.x2 | medium.x3 | large.x1 | large.x2 | large.x3 | large.x4 |
1 KB | No | No | 1700 | 3400 | 6800 | 13600 | 28000 | 56000 | 84000 | 112000 | 224000 | 336000 | 448000 |
1 KB | Yes | No | 500 | 1000 | 2000 | 4000 | 8700 | 17400 | 26100 | 34800 | 69600 | 104400 | 139200 |
Long connections
Response size | HTTPS | gzip | dev.x1 | small.x1 | small.x2 | small.x4 | medium.x1 | medium.x2 | medium.x3 | large.x1 | large.x2 | large.x3 | large.x4 |
1 KB | No | No | 2200 | 4400 | 8800 | 17600 | 35000 | 70000 | 105000 | 140000 | 280000 | 420000 | 560000 |
1 KB | Yes | No | 2000 | 4000 | 8000 | 16000 | 32000 | 64000 | 96000 | 128000 | 256000 | 384000 | 512000 |
1 KB | Yes | Yes | 1700 | 3400 | 6800 | 13600 | 28000 | 56000 | 84000 | 112000 | 224000 | 336000 | 448000 |
10 KB | No | No | 1800 | 3600 | 7200 | 14400 | 30000 | 60000 | 90000 | 120000 | 240000 | 360000 | 480000 |
10 KB | Yes | No | 1700 | 3400 | 6800 | 13600 | 28000 | 56000 | 84000 | 112000 | 224000 | 336000 | 448000 |
10 KB | Yes | Yes | 1000 | 2000 | 4000 | 8000 | 16000 | 32000 | 48000 | 64000 | 128000 | 192000 | 256000 |
Quota limits
Global quotas
Global quotas apply to both Serverless instances and Dedicated instances regardless of specification. To increase quotas, submit a ticket.
Quota item | Quota |
Instances per region | 100 |
Consumer groups per region (max) | 1000 |
Consumers per consumer group | 500 |
Consumers per region (max) | 20000 |
Instance-specific quotas
The table below integrates Serverless per-tenant quotas with Dedicated instance quotas across 11 specifications for side-by-side comparison. Dedicated instances can scale up by upgrading specifications; if the highest specification is still insufficient, submit a ticket for further quota increases.
Serverless | DEV | Small | Medium | Large | ||||||||
Quota item | Serverless | dev.x1 | small.x1 | small.x2 | small.x4 | medium.x1 | medium.x2 | medium.x3 | large.x1 | large.x2 | large.x3 | large.x4 |
Published domains | 100 | 100 | 200 | 300 | 400 | 500 | 600 | 800 | 1000 | 1200 | 1400 | 1600 |
Services | 200 | 200 | 800 | 1200 | 1600 | 2000 | 2500 | 3000 | 4000 | 4500 | 5000 | 6000 |
Online routes total (HTTP API routes, REST API endpoints, Ingress routes) | 2500 | 2500 | 5000 | 6000 | 7000 | 7500 | 8000 | 9000 | 10000 | 11000 | 12000 | 13000 |
Consumer (group) authorization rules | 500 | 500 | 2000 | 4000 | 6000 | 8000 | 12000 | 16000 | 20000 | 27500 | 35000 | 42500 |
Consumer (group) quota rules | 500 | 500 | 2000 | 4000 | 6000 | 8000 | 12000 | 16000 | 20000 | 27500 | 35000 | 42500 |
K8s service sources | 3 | 3 | 3 | 3 | 3 | 5 | 5 | 5 | 5 | 5 | 5 | 5 |
Installed plugins | 5 | 5 | 10 | 10 | 10 | 20 | 20 | 20 | 30 | 30 | 30 | 30 |
Custom plugins uploaded | Not supported | 10 | 20 | 20 | 20 | 50 | 50 | 50 | 80 | 80 | 80 | 80 |
Serverless instance capacity thresholds
Serverless instances use a platform-managed multi-tenant shared pool with no specification choice. All tenants share a unified capacity boundary. The following are per-tenant thresholds for the public beta period.
Per-tenant capacity thresholds
Metric | Beta quota | Description |
Per-tenant QPS | 10000 | Evaluate upgrading to a Dedicated instance when approaching this limit; submit a ticket to exceed it |
New connections per second | 2000 | Platform-enforced per-tenant connection rate limit |
Total connections | 100000 | Per-tenant active connection limit at the gateway ingress |
Elastic capacity
Platform-managed elasticity: Serverless compute resources are auto-scaled by the platform. Tenants do not need to manually scale up or down.
Tenant-level isolation: Each tenant has a dedicated ingress, with traffic strictly isolated from other tenants.
Over-limit handling: When QPS, connections, or routes exceed quota limits, the platform returns throttling responses. For higher quotas, upgrade to a Dedicated instance or submit a ticket.
Quota increases and instance upgrades
Scenario | Recommended path |
Dedicated instance quota insufficient | Upgrade to a higher specification; submit a ticket if still insufficient |
Serverless instance QPS approaching 10000 | Evaluate upgrading to a Dedicated instance |
Serverless instance requires higher SLA | Upgrade to a Dedicated instance |