1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
|
package retry
import (
"math"
"sync"
"time"
"github.com/aws/aws-sdk-go-v2/internal/sdk"
)
type adaptiveRateLimit struct {
tokenBucketEnabled bool
smooth float64
beta float64
scaleConstant float64
minFillRate float64
fillRate float64
calculatedRate float64
lastRefilled time.Time
measuredTxRate float64
lastTxRateBucket float64
requestCount int64
lastMaxRate float64
lastThrottleTime time.Time
timeWindow float64
tokenBucket *adaptiveTokenBucket
mu sync.Mutex
}
func newAdaptiveRateLimit() *adaptiveRateLimit {
now := sdk.NowTime()
return &adaptiveRateLimit{
smooth: 0.8,
beta: 0.7,
scaleConstant: 0.4,
minFillRate: 0.5,
lastTxRateBucket: math.Floor(timeFloat64Seconds(now)),
lastThrottleTime: now,
tokenBucket: newAdaptiveTokenBucket(0),
}
}
func (a *adaptiveRateLimit) Enable(v bool) {
a.mu.Lock()
defer a.mu.Unlock()
a.tokenBucketEnabled = v
}
func (a *adaptiveRateLimit) AcquireToken(amount uint) (
tokenAcquired bool, waitTryAgain time.Duration,
) {
a.mu.Lock()
defer a.mu.Unlock()
if !a.tokenBucketEnabled {
return true, 0
}
a.tokenBucketRefill()
available, ok := a.tokenBucket.Retrieve(float64(amount))
if !ok {
waitDur := float64Seconds((float64(amount) - available) / a.fillRate)
return false, waitDur
}
return true, 0
}
func (a *adaptiveRateLimit) Update(throttled bool) {
a.mu.Lock()
defer a.mu.Unlock()
a.updateMeasuredRate()
if throttled {
rateToUse := a.measuredTxRate
if a.tokenBucketEnabled {
rateToUse = math.Min(a.measuredTxRate, a.fillRate)
}
a.lastMaxRate = rateToUse
a.calculateTimeWindow()
a.lastThrottleTime = sdk.NowTime()
a.calculatedRate = a.cubicThrottle(rateToUse)
a.tokenBucketEnabled = true
} else {
a.calculateTimeWindow()
a.calculatedRate = a.cubicSuccess(sdk.NowTime())
}
newRate := math.Min(a.calculatedRate, 2*a.measuredTxRate)
a.tokenBucketUpdateRate(newRate)
}
func (a *adaptiveRateLimit) cubicSuccess(t time.Time) float64 {
dt := secondsFloat64(t.Sub(a.lastThrottleTime))
return (a.scaleConstant * math.Pow(dt-a.timeWindow, 3)) + a.lastMaxRate
}
func (a *adaptiveRateLimit) cubicThrottle(rateToUse float64) float64 {
return rateToUse * a.beta
}
func (a *adaptiveRateLimit) calculateTimeWindow() {
a.timeWindow = math.Pow((a.lastMaxRate*(1.-a.beta))/a.scaleConstant, 1./3.)
}
func (a *adaptiveRateLimit) tokenBucketUpdateRate(newRPS float64) {
a.tokenBucketRefill()
a.fillRate = math.Max(newRPS, a.minFillRate)
a.tokenBucket.Resize(newRPS)
}
func (a *adaptiveRateLimit) updateMeasuredRate() {
now := sdk.NowTime()
timeBucket := math.Floor(timeFloat64Seconds(now)*2.) / 2.
a.requestCount++
if timeBucket > a.lastTxRateBucket {
currentRate := float64(a.requestCount) / (timeBucket - a.lastTxRateBucket)
a.measuredTxRate = (currentRate * a.smooth) + (a.measuredTxRate * (1. - a.smooth))
a.requestCount = 0
a.lastTxRateBucket = timeBucket
}
}
func (a *adaptiveRateLimit) tokenBucketRefill() {
now := sdk.NowTime()
if a.lastRefilled.IsZero() {
a.lastRefilled = now
return
}
fillAmount := secondsFloat64(now.Sub(a.lastRefilled)) * a.fillRate
a.tokenBucket.Refund(fillAmount)
a.lastRefilled = now
}
func float64Seconds(v float64) time.Duration {
return time.Duration(v * float64(time.Second))
}
func secondsFloat64(v time.Duration) float64 {
return float64(v) / float64(time.Second)
}
func timeFloat64Seconds(v time.Time) float64 {
return float64(v.UnixNano()) / float64(time.Second)
}
|
