1# Ristretto
2[](http://godoc.org/github.com/dgraph-io/ristretto)
3[](https://goreportcard.com/report/github.com/dgraph-io/ristretto)
4[](https://gocover.io/github.com/dgraph-io/ristretto)
5
6
7Ristretto is a fast, concurrent cache library built with a focus on performance and correctness.
8
9The motivation to build Ristretto comes from the need for a contention-free
10cache in [Dgraph][].
11
12[Dgraph]: https://github.com/dgraph-io/dgraph
13
14## Features
15
16* **High Hit Ratios** - with our unique admission/eviction policy pairing, Ristretto's performance is best in class.
17 * **Eviction: SampledLFU** - on par with exact LRU and better performance on Search and Database traces.
18 * **Admission: TinyLFU** - extra performance with little memory overhead (12 bits per counter).
19* **Fast Throughput** - we use a variety of techniques for managing contention and the result is excellent throughput.
20* **Cost-Based Eviction** - any large new item deemed valuable can evict multiple smaller items (cost could be anything).
21* **Fully Concurrent** - you can use as many goroutines as you want with little throughput degradation.
22* **Metrics** - optional performance metrics for throughput, hit ratios, and other stats.
23* **Simple API** - just figure out your ideal `Config` values and you're off and running.
24
25## Status
26
27Ristretto is usable but still under active development. We expect it to be production ready in the near future.
28
29## Table of Contents
30
31* [Usage](#Usage)
32 * [Example](#Example)
33 * [Config](#Config)
34 * [NumCounters](#Config)
35 * [MaxCost](#Config)
36 * [BufferItems](#Config)
37 * [Metrics](#Config)
38 * [OnEvict](#Config)
39 * [KeyToHash](#Config)
40 * [Cost](#Config)
41* [Benchmarks](#Benchmarks)
42 * [Hit Ratios](#Hit-Ratios)
43 * [Search](#Search)
44 * [Database](#Database)
45 * [Looping](#Looping)
46 * [CODASYL](#CODASYL)
47 * [Throughput](#Throughput)
48 * [Mixed](#Mixed)
49 * [Read](#Read)
50 * [Write](#Write)
51* [FAQ](#FAQ)
52
53## Usage
54
55### Example
56
57```go
58func main() {
59 cache, err := ristretto.NewCache(&ristretto.Config{
60 NumCounters: 1e7, // number of keys to track frequency of (10M).
61 MaxCost: 1 << 30, // maximum cost of cache (1GB).
62 BufferItems: 64, // number of keys per Get buffer.
63 })
64 if err != nil {
65 panic(err)
66 }
67
68 // set a value with a cost of 1
69 cache.Set("key", "value", 1)
70
71 // wait for value to pass through buffers
72 time.Sleep(10 * time.Millisecond)
73
74 value, found := cache.Get("key")
75 if !found {
76 panic("missing value")
77 }
78 fmt.Println(value)
79 cache.Del("key")
80}
81```
82
83### Config
84
85The `Config` struct is passed to `NewCache` when creating Ristretto instances (see the example above).
86
87**NumCounters** `int64`
88
89NumCounters is the number of 4-bit access counters to keep for admission and eviction. We've seen good performance in setting this to 10x the number of items you expect to keep in the cache when full.
90
91For example, if you expect each item to have a cost of 1 and MaxCost is 100, set NumCounters to 1,000. Or, if you use variable cost values but expect the cache to hold around 10,000 items when full, set NumCounters to 100,000. The important thing is the *number of unique items* in the full cache, not necessarily the MaxCost value.
92
93**MaxCost** `int64`
94
95MaxCost is how eviction decisions are made. For example, if MaxCost is 100 and a new item with a cost of 1 increases total cache cost to 101, 1 item will be evicted.
96
97MaxCost can also be used to denote the max size in bytes. For example, if MaxCost is 1,000,000 (1MB) and the cache is full with 1,000 1KB items, a new item (that's accepted) would cause 5 1KB items to be evicted.
98
99MaxCost could be anything as long as it matches how you're using the cost values when calling Set.
100
101**BufferItems** `int64`
102
103BufferItems is the size of the Get buffers. The best value we've found for this is 64.
104
105If for some reason you see Get performance decreasing with lots of contention (you shouldn't), try increasing this value in increments of 64. This is a fine-tuning mechanism and you probably won't have to touch this.
106
107**Metrics** `bool`
108
109Metrics is true when you want real-time logging of a variety of stats. The reason this is a Config flag is because there's a 10% throughput performance overhead.
110
111**OnEvict** `func(hashes [2]uint64, value interface{}, cost int64)`
112
113OnEvict is called for every eviction.
114
115**KeyToHash** `func(key interface{}) [2]uint64`
116
117KeyToHash is the hashing algorithm used for every key. If this is nil, Ristretto has a variety of [defaults depending on the underlying interface type](https://github.com/dgraph-io/ristretto/blob/master/z/z.go#L19-L41).
118
119Note that if you want 128bit hashes you should use the full `[2]uint64`,
120otherwise just fill the `uint64` at the `0` position and it will behave like
121any 64bit hash.
122
123**Cost** `func(value interface{}) int64`
124
125Cost is an optional function you can pass to the Config in order to evaluate
126item cost at runtime, and only for the Set calls that aren't dropped (this is
127useful if calculating item cost is particularly expensive and you don't want to
128waste time on items that will be dropped anyways).
129
130To signal to Ristretto that you'd like to use this Cost function:
131
1321. Set the Cost field to a non-nil function.
1332. When calling Set for new items or item updates, use a `cost` of 0.
134
135## Benchmarks
136
137The benchmarks can be found in https://github.com/dgraph-io/benchmarks/tree/master/cachebench/ristretto.
138
139### Hit Ratios
140
141#### Search
142
143This trace is described as "disk read accesses initiated by a large commercial
144search engine in response to various web search requests."
145
146<p align="center">
147 <img src="https://raw.githubusercontent.com/dgraph-io/ristretto/master/benchmarks/Hit%20Ratios%20-%20Search%20(ARC-S3).svg">
148</p>
149
150#### Database
151
152This trace is described as "a database server running at a commercial site
153running an ERP application on top of a commercial database."
154
155<p align="center">
156 <img src="https://raw.githubusercontent.com/dgraph-io/ristretto/master/benchmarks/Hit%20Ratios%20-%20Database%20(ARC-DS1).svg">
157</p>
158
159#### Looping
160
161This trace demonstrates a looping access pattern.
162
163<p align="center">
164 <img src="https://raw.githubusercontent.com/dgraph-io/ristretto/master/benchmarks/Hit%20Ratios%20-%20Glimpse%20(LIRS-GLI).svg">
165</p>
166
167#### CODASYL
168
169This trace is described as "references to a CODASYL database for a one hour
170period."
171
172<p align="center">
173 <img src="https://raw.githubusercontent.com/dgraph-io/ristretto/master/benchmarks/Hit%20Ratios%20-%20CODASYL%20(ARC-OLTP).svg">
174</p>
175
176### Throughput
177
178All throughput benchmarks were ran on an Intel Core i7-8700K (3.7GHz) with 16gb
179of RAM.
180
181#### Mixed
182
183<p align="center">
184 <img src="https://raw.githubusercontent.com/dgraph-io/ristretto/master/benchmarks/Throughput%20-%20Mixed.svg">
185</p>
186
187#### Read
188
189<p align="center">
190 <img src="https://raw.githubusercontent.com/dgraph-io/ristretto/master/benchmarks/Throughput%20-%20Read%20(Zipfian).svg">
191</p>
192
193#### Write
194
195<p align="center">
196 <img src="https://raw.githubusercontent.com/dgraph-io/ristretto/master/benchmarks/Throughput%20-%20Write%20(Zipfian).svg">
197</p>
198
199## FAQ
200
201### How are you achieving this performance? What shortcuts are you taking?
202
203We go into detail in the [Ristretto blog post](https://blog.dgraph.io/post/introducing-ristretto-high-perf-go-cache/), but in short: our throughput performance can be attributed to a mix of batching and eventual consistency. Our hit ratio performance is mostly due to an excellent [admission policy](https://arxiv.org/abs/1512.00727) and SampledLFU eviction policy.
204
205As for "shortcuts," the only thing Ristretto does that could be construed as one is dropping some Set calls. That means a Set call for a new item (updates are guaranteed) isn't guaranteed to make it into the cache. The new item could be dropped at two points: when passing through the Set buffer or when passing through the admission policy. However, this doesn't affect hit ratios much at all as we expect the most popular items to be Set multiple times and eventually make it in the cache.
206
207### Is Ristretto distributed?
208
209No, it's just like any other Go library that you can import into your project and use in a single process.
View as plain text