Files
easyai-ai-gateway/apps/api/internal/acceptanceworkload/profiles.go
T
wangbo c28bf74230 feat(worker): 实现集群限流与自适应负载
保留平台模型 RPM、TPM 和并发策略语义,增加 PostgreSQL 集群级租约、饱和候选重选和多平台自动负载,避免突发任务固定等待首个平台。\n\n新增 Worker 实时负载采样、自适应 active/heavy 容量、心跳与管理端指标,并扩展本地 acceptance runner,覆盖三 Worker、同模型三平台 2/4/6 并发和 48 个带图视频突发任务。\n\n验证:go test ./...、go vet ./...、PostgreSQL 跨 Store 集成测试、gofmt、bash -n、ShellCheck 及本地集群 provider-burst 验收通过;48/48 成功,无越限、重复提交、重复计费、重复回调或终态资源泄漏。
2026-08-03 00:13:46 +08:00

60 lines
1.5 KiB
Go

package acceptanceworkload
import "time"
type GeminiProfile struct {
Name string
Requests int
InputImages int
InputBytes int
OutputBytes int
DelayMin time.Duration
DelayMax time.Duration
}
var (
GeminiBaseline = GeminiProfile{
Name: "gemini-baseline", Requests: 1000, InputImages: 1, InputBytes: 256 << 10, OutputBytes: 256 << 10,
DelayMin: 4 * time.Second, DelayMax: 4 * time.Second,
}
GeminiMultiImage = GeminiProfile{
Name: "gemini-multi-image", Requests: 192, InputImages: 3, InputBytes: 768 << 10, OutputBytes: 0,
DelayMin: 8 * time.Second, DelayMax: 15 * time.Second,
}
GeminiLarge = GeminiProfile{
Name: "gemini-large", Requests: 128, InputImages: 1, InputBytes: 2 << 20, OutputBytes: 4 << 20,
DelayMin: 8 * time.Second, DelayMax: 15 * time.Second,
}
GeminiPeak = GeminiProfile{
Name: "gemini-peak", Requests: 32, InputImages: 1, InputBytes: 8 << 20, OutputBytes: 8 << 20,
DelayMin: 15 * time.Second, DelayMax: 30 * time.Second,
}
)
func GeminiProfileByName(name string) (GeminiProfile, bool) {
switch name {
case GeminiBaseline.Name:
return GeminiBaseline, true
case GeminiMultiImage.Name:
return GeminiMultiImage, true
case GeminiLarge.Name:
return GeminiLarge, true
case GeminiPeak.Name:
return GeminiPeak, true
default:
return GeminiProfile{}, false
}
}
// VideoImageCount provides a deterministic 60/30/10 split for 3/6/9-image tasks.
func VideoImageCount(index int) int {
switch index % 10 {
case 0, 1, 2, 3, 4, 5:
return 3
case 6, 7, 8:
return 6
default:
return 9
}
}