Expand description

Stretto

Stretto is a pure Rust implementation for https://github.com/dgraph-io/ristretto.

A high performance thread-safe memory-bound Rust cache.

English | 简体中文

github Build codecov

docs.rs crates.io rustc

license-apache license-mit

Features

  • Internal Mutability - Do not need to use Arc<RwLock<Cache<...>> for concurrent code, you just need Cache<...> or AsyncCache<...>
  • Sync and Async - Stretto support sync and runtime agnostic async.
    • In sync, Cache starts two extra OS level threads. One is policy thread, the other is writing thread.
    • In async, AsyncCache starts two extra green threads. One is policy thread, the other is writing thread.
  • Store policy Stretto only store the value, which means the cache does not store the key.
  • High Hit Ratios - with Dgrpah’s developers unique admission/eviction policy pairing, Stretto’s performance is best in class.
  • Eviction: SampledLFU - on par with exact LRU and better performance on Search and Database traces.
  • Admission: TinyLFU - extra performance with little memory overhead (12 bits per counter).
  • Fast Throughput - use a variety of techniques for managing contention and the result is excellent throughput.
  • Cost-Based Eviction - any large new item deemed valuable can evict multiple smaller items (cost could be anything).
  • Fully Concurrent - you can use as many threads as you want with little throughput degradation.
  • Metrics - optional performance metrics for throughput, hit ratios, and other stats.
  • Simple API - just figure out your ideal CacheBuilder values and you’re off and running.

Table of Contents

Usage

Example

Please see examples on github.

Config

The CacheBuilder or AsyncCacheBuilder struct is used when creating Cache/AsyncCache instances if you want to customize the Cache/AsyncCache settings.

num_counters

num_counters is the number of 4-bit access counters to keep for admission and eviction. Dgraph’s developers have seen good performance in setting this to 10x the number of items you expect to keep in the cache when full.

For example, if you expect each item to have a cost of 1 and max_cost is 100, set num_counters to 1,000. Or, if you use variable cost values but expect the cache to hold around 10,000 items when full, set num_counters to 100,000. The important thing is the number of unique items in the full cache, not necessarily the max_cost value.

max_cost

max_cost is how eviction decisions are made. For example, if max_cost is 100 and a new item with a cost of 1 increases total cache cost to 101, 1 item will be evicted.

max_cost can also be used to denote the max size in bytes. For example, if max_cost 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.

max_cost could be anything as long as it matches how you’re using the cost values when calling insert.

key_builder

KeyBuilder is the hashing algorithm used for every key. In Stretto, the Cache will never store the real key. The key will be processed by KeyBuilder. Stretto has two default built-in key builder, one is TransparentKeyBuilder, the other is DefaultKeyBuilder. If your key implements TransparentKey trait, you can use TransparentKeyBuilder which is faster than DefaultKeyBuilder. Otherwise, you should use DefaultKeyBuilder You can also write your own key builder for the Cache, by implementing KeyBuilder trait.

Note that if you want 128bit hashes you should use the full (u64, u64), otherwise just fill the u64 at the 0 position, and it will behave like any 64bit hash.

buffer_size

buffer_size is the size of the insert buffers. The Dgraph’s developers find that 32 * 1024 gives a good performance.

If for some reason you see insert performance decreasing with lots of contention (you shouldn’t), try increasing this value in increments of 32 * 1024. This is a fine-tuning mechanism and you probably won’t have to touch this.

metrics

Metrics is true when you want real-time logging of a variety of stats. The reason this is a CacheBuilder flag is because there’s a 10% throughput performance overhead.

ignore_internal_cost

Set to true indicates to the cache that the cost of internally storing the value should be ignored. This is useful when the cost passed to set is not using bytes as units. Keep in mind that setting this to true will increase the memory usage.

cleanup_duration

The Cache will cleanup the expired values every 500ms by default.

update_validator

By default, the Cache will always update the value if the value already exists in the cache. UpdateValidator is a trait to support customized update policy (check if the value should be updated if the value already exists in the cache).

callback

CacheCallback is for customize some extra operations on values when related event happens.

coster

Coster is a trait you can pass to the CacheBuilder in order to evaluate item cost at runtime, and only for the insert calls that aren’t dropped (this is useful if calculating item cost is particularly expensive, and you don’t want to waste time on items that will be dropped anyways).

To signal to Stretto that you’d like to use this Coster trait:

  1. Set the Coster field to your own Coster implementation.
  2. When calling insert for new items or item updates, use a cost of 0.
hasher

The hasher for the Cache, default is SipHasher.

Structs

AsyncCache is a thread-safe async implementation of a hashmap with a TinyLFU admission policy and a Sampled LFU eviction policy. You can use the same AsyncCache instance from as many threads as you want.
The AsyncCacheBuilder struct is used when creating AsyncCache instances if you want to customize the AsyncCache settings.
Cachesync
Cache is a thread-safe implementation of a hashmap with a TinyLFU admission policy and a Sampled LFU eviction policy. You can use the same Cache instance from as many threads as you want.
The CacheBuilder struct is used when creating Cache instances if you want to customize the Cache settings.
DefaultKeyBuilder is a built-in KeyBuilder for the Cache.
Histogram stores the information needed to represent the sizes of the keys and values as a histogram.
Item is the parameter when Cache reject, evict value,
TransparentKeyBuilder converts key to u64. If the key does not implement the trait TransparentKey, please use DefaultKeyBuilder or write a custom key builder.
ValueRef is returned when invoking get method of the Cache. It contains a RwLockReadGuard and a value reference.
ValueRefMut is returned when invoking get_mut method of the Cache. It contains a RwLockWriteGuard and a mutable value reference.

Enums

CacheError contains the error of this crate
The data field in a Metrics
Metrics is a snapshot of performance statistics for the lifetime of a cache instance.

Traits

CacheCallback is for customize some extra operations on values when related event happens.
Cost is a trait you can pass to the CacheBuilder in order to evaluate item cost at runtime, and only for the insert calls that aren’t dropped (this is useful if calculating item cost is particularly expensive, and you don’t want to waste time on items that will be dropped anyways).
KeyBuilder is the hashing algorithm used for every key. In Stretto, the Cache will never store the real key. The key will be processed by KeyBuilder. Stretto has two default built-in key builder, one is TransparentKeyBuilder, the other is DefaultKeyBuilder. If your key implements TransparentKey trait, you can use TransparentKeyBuilder which is faster than DefaultKeyBuilder. Otherwise, you should use DefaultKeyBuilder You can also write your own key builder for the Cache, by implementing KeyBuilder trait.
Implement this trait for the key, if you want to use TransparentKeyBuilder as the KeyBuilder for the Cache.
By default, the Cache will always update the value if the value already exists in the cache, this trait is for you to check if the value should be updated.