多圖剖析緩存之王 Caffeine 高性能設(shè)計(jì)
以下文章來(lái)源于武培軒,回復(fù)資源獲取資料
概要
Caffeine[1]是一個(gè)高性能,高命中率,低內(nèi)存占用,near optimal 的本地緩存,簡(jiǎn)單來(lái)說(shuō)它是 Guava Cache 的優(yōu)化加強(qiáng)版,有些文章把 Caffeine 稱(chēng)為“新一代的緩存”、“現(xiàn)代緩存之王”。
本文將重點(diǎn)講解 Caffeine 的高性能設(shè)計(jì),以及對(duì)應(yīng)部分的源碼分析。
與 Guava Cache 比較
如果你對(duì) Guava Cache 還不理解的話,可以點(diǎn)擊這里[2]來(lái)看一下我之前寫(xiě)過(guò)關(guān)于 Guava Cache 的文章。
大家都知道,Spring5 即將放棄掉 Guava Cache 作為緩存機(jī)制,而改用 Caffeine 作為新的本地 Cache 的組件,這對(duì)于 Caffeine 來(lái)說(shuō)是一個(gè)很大的肯定。為什么 Spring 會(huì)這樣做呢?其實(shí)在 Caffeine 的Benchmarks[3]里給出了好靚仔的數(shù)據(jù),對(duì)讀和寫(xiě)的場(chǎng)景,還有跟其他幾個(gè)緩存工具進(jìn)行了比較,Caffeine 的性能都表現(xiàn)很突出。

使用 Caffeine
Caffeine 為了方便大家使用以及從 Guava Cache 切換過(guò)來(lái)(很有針對(duì)性啊~),借鑒了 Guava Cache 大部分的概念(諸如核心概念Cache、LoadingCache、CacheLoader、CacheBuilder等等),對(duì)于 Caffeine 的理解只要把它當(dāng)作 Guava Cache 就可以了。
使用上,大家只要把 Caffeine 的包引進(jìn)來(lái),然后換一下 cache 的實(shí)現(xiàn)類(lèi),基本應(yīng)該就沒(méi)問(wèn)題了。這對(duì)與已經(jīng)使用過(guò) Guava Cache 的同學(xué)來(lái)說(shuō)沒(méi)有任何難度,甚至還有一點(diǎn)熟悉的味道,如果你之前沒(méi)有使用過(guò) Guava Cache,可以查看 Caffeine 的官方 API 說(shuō)明文檔[4],其中Population,Eviction,Removal,Refresh,Statistics,Cleanup,Policy等等這些特性都是跟 Guava Cache 基本一樣的。
下面給出一個(gè)例子說(shuō)明怎樣創(chuàng)建一個(gè) Cache:
private static LoadingCache<String, String> cache = Caffeine.newBuilder()
//最大個(gè)數(shù)限制
.maximumSize(256L)
//初始化容量
.initialCapacity(1)
//訪問(wèn)后過(guò)期(包括讀和寫(xiě))
.expireAfterAccess(2, TimeUnit.DAYS)
//寫(xiě)后過(guò)期
.expireAfterWrite(2, TimeUnit.HOURS)
//寫(xiě)后自動(dòng)異步刷新
.refreshAfterWrite(1, TimeUnit.HOURS)
//記錄下緩存的一些統(tǒng)計(jì)數(shù)據(jù),例如命中率等
.recordStats()
//cache對(duì)緩存寫(xiě)的通知回調(diào)
.writer(new CacheWriter<Object, Object>() {
@Override
public void write(@NonNull Object key, @NonNull Object value) {
log.info("key={}, CacheWriter write", key);
}
@Override
public void delete(@NonNull Object key, @Nullable Object value, @NonNull RemovalCause cause) {
log.info("key={}, cause={}, CacheWriter delete", key, cause);
}
})
//使用CacheLoader創(chuàng)建一個(gè)LoadingCache
.build(new CacheLoader<String, String>() {
//同步加載數(shù)據(jù)
@Nullable
@Override
public String load(@NonNull String key) throws Exception {
return "value_" + key;
}
//異步加載數(shù)據(jù)
@Nullable
@Override
public String reload(@NonNull String key, @NonNull String oldValue) throws Exception {
return "value_" + key;
}
});
更多從 Guava Cache 遷移過(guò)來(lái)的使用說(shuō)明,請(qǐng)看這里[5]
Caffeine 的高性能設(shè)計(jì)
判斷一個(gè)緩存的好壞最核心的指標(biāo)就是命中率,影響緩存命中率有很多因素,包括業(yè)務(wù)場(chǎng)景、淘汰策略、清理策略、緩存容量等等。如果作為本地緩存, 它的性能的情況,資源的占用也都是一個(gè)很重要的指標(biāo)。下面
我們來(lái)看看 Caffeine 在這幾個(gè)方面是怎么著手的,如何做優(yōu)化的。
(注:本文不會(huì)分析 Caffeine 全部源碼,只會(huì)對(duì)核心設(shè)計(jì)的實(shí)現(xiàn)進(jìn)行分析,但我建議讀者把 Caffeine 的源碼都涉獵一下,有個(gè) overview 才能更好理解本文。如果你看過(guò) Guava Cache 的源碼也行,代碼的數(shù)據(jù)結(jié)構(gòu)和處理邏輯很類(lèi)似的。
源碼基于:caffeine-2.8.0.jar)
W-TinyLFU 整體設(shè)計(jì)
上面說(shuō)到淘汰策略是影響緩存命中率的因素之一,一般比較簡(jiǎn)單的緩存就會(huì)直接用到 LFU(Least Frequently Used,即最不經(jīng)常使用) 或者LRU(Least Recently Used,即最近最少使用) ,而 Caffeine 就是使用了 W-TinyLFU 算法。
W-TinyLFU 看名字就能大概猜出來(lái),它是 LFU 的變種,也是一種緩存淘汰算法。那為什么要使用 W-TinyLFU 呢?
LRU 和 LFU 的缺點(diǎn)
LRU 實(shí)現(xiàn)簡(jiǎn)單,在一般情況下能夠表現(xiàn)出很好的命中率,是一個(gè)“性?xún)r(jià)比”很高的算法,平時(shí)也很常用。雖然 LRU 對(duì)突發(fā)性的稀疏流量(sparse bursts)表現(xiàn)很好,但同時(shí)也會(huì)產(chǎn)生緩存污染,舉例來(lái)說(shuō),如果偶然性的要對(duì)全量數(shù)據(jù)進(jìn)行遍歷,那么“歷史訪問(wèn)記錄”就會(huì)被刷走,造成污染。 如果數(shù)據(jù)的分布在一段時(shí)間內(nèi)是固定的話,那么 LFU 可以達(dá)到最高的命中率。但是 LFU 有兩個(gè)缺點(diǎn),第一,它需要給每個(gè)記錄項(xiàng)維護(hù)頻率信息,每次訪問(wèn)都需要更新,這是個(gè)巨大的開(kāi)銷(xiāo);第二,對(duì)突發(fā)性的稀疏流量無(wú)力,因?yàn)榍捌诮?jīng)常訪問(wèn)的記錄已經(jīng)占用了緩存,偶然的流量不太可能會(huì)被保留下來(lái),而且過(guò)去的一些大量被訪問(wèn)的記錄在將來(lái)也不一定會(huì)使用上,這樣就一直把“坑”占著了。
無(wú)論 LRU 還是 LFU 都有其各自的缺點(diǎn),不過(guò),現(xiàn)在已經(jīng)有很多針對(duì)其缺點(diǎn)而改良、優(yōu)化出來(lái)的變種算法。
TinyLFU
TinyLFU 就是其中一個(gè)優(yōu)化算法,它是專(zhuān)門(mén)為了解決 LFU 上述提到的兩個(gè)問(wèn)題而被設(shè)計(jì)出來(lái)的。
解決第一個(gè)問(wèn)題是采用了 Count–Min Sketch 算法。
解決第二個(gè)問(wèn)題是讓記錄盡量保持相對(duì)的“新鮮”(Freshness Mechanism),并且當(dāng)有新的記錄插入時(shí),可以讓它跟老的記錄進(jìn)行“PK”,輸者就會(huì)被淘汰,這樣一些老的、不再需要的記錄就會(huì)被剔除。
下圖是 TinyLFU 設(shè)計(jì)圖(來(lái)自官方)

統(tǒng)計(jì)頻率 Count–Min Sketch 算法
如何對(duì)一個(gè) key 進(jìn)行統(tǒng)計(jì),但又可以節(jié)省空間呢?(不是簡(jiǎn)單的使用HashMap,這太消耗內(nèi)存了),注意哦,不需要精確的統(tǒng)計(jì),只需要一個(gè)近似值就可以了,怎么樣,這樣場(chǎng)景是不是很熟悉,如果你是老司機(jī),或許已經(jīng)聯(lián)想到布隆過(guò)濾器(Bloom Filter)的應(yīng)用了。
沒(méi)錯(cuò),將要介紹的 Count–Min Sketch 的原理跟 Bloom Filter 一樣,只不過(guò) Bloom Filter 只有 0 和 1 的值,那么你可以把 Count–Min Sketch 看作是“數(shù)值”版的 Bloom Filter。
更多關(guān)于 Count–Min Sketch 的介紹請(qǐng)自行搜索。
在 TinyLFU 中,近似頻率的統(tǒng)計(jì)如下圖所示:

對(duì)一個(gè) key 進(jìn)行多次 hash 函數(shù)后,index 到多個(gè)數(shù)組位置后進(jìn)行累加,查詢(xún)時(shí)取多個(gè)值中的最小值即可。
Caffeine 對(duì)這個(gè)算法的實(shí)現(xiàn)在FrequencySketch類(lèi)。但 Caffeine 對(duì)此有進(jìn)一步的優(yōu)化,例如 Count–Min Sketch 使用了二維數(shù)組,Caffeine 只是用了一個(gè)一維的數(shù)組;再者,如果是數(shù)值類(lèi)型的話,這個(gè)數(shù)需要用 int 或 long 來(lái)存儲(chǔ),但是 Caffeine 認(rèn)為緩存的訪問(wèn)頻率不需要用到那么大,只需要 15 就足夠,一般認(rèn)為達(dá)到 15 次的頻率算是很高的了,而且 Caffeine 還有另外一個(gè)機(jī)制來(lái)使得這個(gè)頻率進(jìn)行衰退減半(下面就會(huì)講到)。如果最大是 15 的話,那么只需要 4 個(gè) bit 就可以滿(mǎn)足了,一個(gè) long 有 64bit,可以存儲(chǔ) 16 個(gè)這樣的統(tǒng)計(jì)數(shù),Caffeine 就是這樣的設(shè)計(jì),使得存儲(chǔ)效率提高了 16 倍。
Caffeine 對(duì)緩存的讀寫(xiě)(afterRead和afterWrite方法)都會(huì)調(diào)用onAccesss 方法,而onAccess方法里有一句:
frequencySketch().increment(key);
這句就是追加記錄的頻率,下面我們看看具體實(shí)現(xiàn)
//FrequencySketch的一些屬性
//種子數(shù)
static final long[] SEED = { // A mixture of seeds from FNV-1a, CityHash, and Murmur3
0xc3a5c85c97cb3127L, 0xb492b66fbe98f273L, 0x9ae16a3b2f90404fL, 0xcbf29ce484222325L};
static final long RESET_MASK = 0x7777777777777777L;
static final long ONE_MASK = 0x1111111111111111L;
int sampleSize;
//為了快速根據(jù)hash值得到table的index值的掩碼
//table的長(zhǎng)度size一般為2的n次方,而tableMask為size-1,這樣就可以通過(guò)&操作來(lái)模擬取余操作,速度快很多,老司機(jī)都知道
int tableMask;
//存儲(chǔ)數(shù)據(jù)的一維long數(shù)組
long[] table;
int size;
/**
* Increments the popularity of the element if it does not exceed the maximum (15). The popularity
* of all elements will be periodically down sampled when the observed events exceeds a threshold.
* This process provides a frequency aging to allow expired long term entries to fade away.
*
* @param e the element to add
*/
public void increment(@NonNull E e) {
if (isNotInitialized()) {
return;
}
//根據(jù)key的hashCode通過(guò)一個(gè)哈希函數(shù)得到一個(gè)hash值
//本來(lái)就是hashCode了,為什么還要再做一次hash?怕原來(lái)的hashCode不夠均勻分散,再打散一下。
int hash = spread(e.hashCode());
//這句光看有點(diǎn)難理解
//就如我剛才說(shuō)的,Caffeine把一個(gè)long的64bit劃分成16個(gè)等分,每一等分4個(gè)bit。
//這個(gè)start就是用來(lái)定位到是哪一個(gè)等分的,用hash值低兩位作為隨機(jī)數(shù),再左移2位,得到一個(gè)小于16的值
int start = (hash & 3) << 2;
//indexOf方法的意思就是,根據(jù)hash值和不同種子得到table的下標(biāo)index
//這里通過(guò)四個(gè)不同的種子,得到四個(gè)不同的下標(biāo)index
int index0 = indexOf(hash, 0);
int index1 = indexOf(hash, 1);
int index2 = indexOf(hash, 2);
int index3 = indexOf(hash, 3);
//根據(jù)index和start(+1, +2, +3)的值,把table[index]對(duì)應(yīng)的等分追加1
//這個(gè)incrementAt方法有點(diǎn)難理解,看我下面的解釋
boolean added = incrementAt(index0, start);
added |= incrementAt(index1, start + 1);
added |= incrementAt(index2, start + 2);
added |= incrementAt(index3, start + 3);
//這個(gè)reset等下說(shuō)
if (added && (++size == sampleSize)) {
reset();
}
}
/**
* Increments the specified counter by 1 if it is not already at the maximum value (15).
*
* @param i the table index (16 counters)
* @param j the counter to increment
* @return if incremented
*/
boolean incrementAt(int i, int j) {
//這個(gè)j表示16個(gè)等分的下標(biāo),那么offset就是相當(dāng)于在64位中的下標(biāo)(這個(gè)自己想想)
int offset = j << 2;
//上面提到Caffeine把頻率統(tǒng)計(jì)最大定為15,即0xfL
//mask就是在64位中的掩碼,即1111后面跟很多個(gè)0
long mask = (0xfL << offset);
//如果&的結(jié)果不等于15,那么就追加1。等于15就不會(huì)再加了
if ((table[i] & mask) != mask) {
table[i] += (1L << offset);
return true;
}
return false;
}
/**
* Returns the table index for the counter at the specified depth.
*
* @param item the element's hash
* @param i the counter depth
* @return the table index
*/
int indexOf(int item, int i) {
long hash = SEED[i] * item;
hash += hash >>> 32;
return ((int) hash) & tableMask;
}
/**
* Applies a supplemental hash function to a given hashCode, which defends against poor quality
* hash functions.
*/
int spread(int x) {
x = ((x >>> 16) ^ x) * 0x45d9f3b;
x = ((x >>> 16) ^ x) * 0x45d9f3b;
return (x >>> 16) ^ x;
}
知道了追加方法,那么讀取方法frequency就很容易理解了。
/**
* Returns the estimated number of occurrences of an element, up to the maximum (15).
*
* @param e the element to count occurrences of
* @return the estimated number of occurrences of the element; possibly zero but never negative
*/
@NonNegative
public int frequency(@NonNull E e) {
if (isNotInitialized()) {
return 0;
}
//得到hash值,跟上面一樣
int hash = spread(e.hashCode());
//得到等分的下標(biāo),跟上面一樣
int start = (hash & 3) << 2;
int frequency = Integer.MAX_VALUE;
//循環(huán)四次,分別獲取在table數(shù)組中不同的下標(biāo)位置
for (int i = 0; i < 4; i++) {
int index = indexOf(hash, i);
//這個(gè)操作就不多說(shuō)了,其實(shí)跟上面incrementAt是一樣的,定位到table[index] + 等分的位置,再根據(jù)mask取出計(jì)數(shù)值
int count = (int) ((table[index] >>> ((start + i) << 2)) & 0xfL);
//取四個(gè)中的較小值
frequency = Math.min(frequency, count);
}
return frequency;
}
通過(guò)代碼和注釋或者讀者可能難以理解,下圖是我畫(huà)出來(lái)幫助大家理解的結(jié)構(gòu)圖。
注意紫色虛線框,其中藍(lán)色小格就是需要計(jì)算的位置:

保新機(jī)制
為了讓緩存保持“新鮮”,剔除掉過(guò)往頻率很高但之后不經(jīng)常的緩存,Caffeine 有一個(gè) Freshness Mechanism。做法很簡(jiǎn)答,就是當(dāng)整體的統(tǒng)計(jì)計(jì)數(shù)(當(dāng)前所有記錄的頻率統(tǒng)計(jì)之和,這個(gè)數(shù)值內(nèi)部維護(hù))達(dá)到某一個(gè)值時(shí),那么所有記錄的頻率統(tǒng)計(jì)除以 2。
從上面的代碼
//size變量就是所有記錄的頻率統(tǒng)計(jì)之,即每個(gè)記錄加1,這個(gè)size都會(huì)加1
//sampleSize一個(gè)閾值,從FrequencySketch初始化可以看到它的值為maximumSize的10倍
if (added && (++size == sampleSize)) {
reset();
}
看到reset方法就是做這個(gè)事情
/** Reduces every counter by half of its original value. */
void reset() {
int count = 0;
for (int i = 0; i < table.length; i++) {
count += Long.bitCount(table[i] & ONE_MASK);
table[i] = (table[i] >>> 1) & RESET_MASK;
}
size = (size >>> 1) - (count >>> 2);
}
關(guān)于這個(gè) reset 方法,為什么是除以 2,而不是其他,及其正確性,在最下面的參考資料的 TinyLFU 論文中 3.3 章節(jié)給出了數(shù)學(xué)證明,大家有興趣可以看看。
增加一個(gè) Window?
Caffeine 通過(guò)測(cè)試發(fā)現(xiàn) TinyLFU 在面對(duì)突發(fā)性的稀疏流量(sparse bursts)時(shí)表現(xiàn)很差,因?yàn)樾碌挠涗洠╪ew items)還沒(méi)來(lái)得及建立足夠的頻率就被剔除出去了,這就使得命中率下降。
于是 Caffeine 設(shè)計(jì)出一種新的 policy,即 Window Tiny LFU(W-TinyLFU),并通過(guò)實(shí)驗(yàn)和實(shí)踐發(fā)現(xiàn) W-TinyLFU 比 TinyLFU 表現(xiàn)的更好。
W-TinyLFU 的設(shè)計(jì)如下所示(兩圖等價(jià)):


它主要包括兩個(gè)緩存模塊,主緩存是 SLRU(Segmented LRU,即分段 LRU),SLRU 包括一個(gè)名為 protected 和一個(gè)名為 probation 的緩存區(qū)。通過(guò)增加一個(gè)緩存區(qū)(即 Window Cache),當(dāng)有新的記錄插入時(shí),會(huì)先在 window 區(qū)呆一下,就可以避免上述說(shuō)的 sparse bursts 問(wèn)題。
淘汰策略(eviction policy)
當(dāng) window 區(qū)滿(mǎn)了,就會(huì)根據(jù) LRU 把 candidate(即淘汰出來(lái)的元素)放到 probation 區(qū),如果 probation 區(qū)也滿(mǎn)了,就把 candidate 和 probation 將要淘汰的元素 victim,兩個(gè)進(jìn)行“PK”,勝者留在 probation,輸者就要被淘汰了。
而且經(jīng)過(guò)實(shí)驗(yàn)發(fā)現(xiàn)當(dāng) window 區(qū)配置為總?cè)萘康?1%,剩余的 99%當(dāng)中的 80%分給 protected 區(qū),20%分給 probation 區(qū)時(shí),這時(shí)整體性能和命中率表現(xiàn)得最好,所以 Caffeine 默認(rèn)的比例設(shè)置就是這個(gè)。
不過(guò)這個(gè)比例 Caffeine 會(huì)在運(yùn)行時(shí)根據(jù)統(tǒng)計(jì)數(shù)據(jù)(statistics)去動(dòng)態(tài)調(diào)整,如果你的應(yīng)用程序的緩存隨著時(shí)間變化比較快的話,那么增加 window 區(qū)的比例可以提高命中率,相反緩存都是比較固定不變的話,增加 Main Cache 區(qū)(protected 區(qū) +probation 區(qū))的比例會(huì)有較好的效果。
下面我們看看上面說(shuō)到的淘汰策略是怎么實(shí)現(xiàn)的:
一般緩存對(duì)讀寫(xiě)操作后都有后續(xù)的一系列“維護(hù)”操作,Caffeine 也不例外,這些操作都在maintenance方法,我們將要說(shuō)到的淘汰策略也在里面。
這方法比較重要,下面也會(huì)提到,所以這里只先說(shuō)跟“淘汰策略”有關(guān)的evictEntries和climb。
/**
* Performs the pending maintenance work and sets the state flags during processing to avoid
* excess scheduling attempts. The read buffer, write buffer, and reference queues are
* drained, followed by expiration, and size-based eviction.
*
* @param task an additional pending task to run, or {@code null} if not present
*/
@GuardedBy("evictionLock")
void maintenance(@Nullable Runnable task) {
lazySetDrainStatus(PROCESSING_TO_IDLE);
try {
drainReadBuffer();
drainWriteBuffer();
if (task != null) {
task.run();
}
drainKeyReferences();
drainValueReferences();
expireEntries();
//把符合條件的記錄淘汰掉
evictEntries();
//動(dòng)態(tài)調(diào)整window區(qū)和protected區(qū)的大小
climb();
} finally {
if ((drainStatus() != PROCESSING_TO_IDLE) || !casDrainStatus(PROCESSING_TO_IDLE, IDLE)) {
lazySetDrainStatus(REQUIRED);
}
}
}
先說(shuō)一下 Caffeine 對(duì)上面說(shuō)到的 W-TinyLFU 策略的實(shí)現(xiàn)用到的數(shù)據(jù)結(jié)構(gòu):
//最大的個(gè)數(shù)限制
long maximum;
//當(dāng)前的個(gè)數(shù)
long weightedSize;
//window區(qū)的最大限制
long windowMaximum;
//window區(qū)當(dāng)前的個(gè)數(shù)
long windowWeightedSize;
//protected區(qū)的最大限制
long mainProtectedMaximum;
//protected區(qū)當(dāng)前的個(gè)數(shù)
long mainProtectedWeightedSize;
//下一次需要調(diào)整的大小(還需要進(jìn)一步計(jì)算)
double stepSize;
//window區(qū)需要調(diào)整的大小
long adjustment;
//命中計(jì)數(shù)
int hitsInSample;
//不命中的計(jì)數(shù)
int missesInSample;
//上一次的緩存命中率
double previousSampleHitRate;
final FrequencySketch<K> sketch;
//window區(qū)的LRU queue(FIFO)
final AccessOrderDeque<Node<K, V>> accessOrderWindowDeque;
//probation區(qū)的LRU queue(FIFO)
final AccessOrderDeque<Node<K, V>> accessOrderProbationDeque;
//protected區(qū)的LRU queue(FIFO)
final AccessOrderDeque<Node<K, V>> accessOrderProtectedDeque;
以及默認(rèn)比例設(shè)置(意思看注釋?zhuān)?/p>
/** The initial percent of the maximum weighted capacity dedicated to the main space. */
static final double PERCENT_MAIN = 0.99d;
/** The percent of the maximum weighted capacity dedicated to the main's protected space. */
static final double PERCENT_MAIN_PROTECTED = 0.80d;
/** The difference in hit rates that restarts the climber. */
static final double HILL_CLIMBER_RESTART_THRESHOLD = 0.05d;
/** The percent of the total size to adapt the window by. */
static final double HILL_CLIMBER_STEP_PERCENT = 0.0625d;
/** The rate to decrease the step size to adapt by. */
static final double HILL_CLIMBER_STEP_DECAY_RATE = 0.98d;
/** The maximum number of entries that can be transfered between queues. */
重點(diǎn)來(lái)了,evictEntries和climb方法:
/** Evicts entries if the cache exceeds the maximum. */
@GuardedBy("evictionLock")
void evictEntries() {
if (!evicts()) {
return;
}
//淘汰window區(qū)的記錄
int candidates = evictFromWindow();
//淘汰Main區(qū)的記錄
evictFromMain(candidates);
}
/**
* Evicts entries from the window space into the main space while the window size exceeds a
* maximum.
*
* @return the number of candidate entries evicted from the window space
*/
//根據(jù)W-TinyLFU,新的數(shù)據(jù)都會(huì)無(wú)條件的加到admission window
//但是window是有大小限制,所以要“定期”做一下“維護(hù)”
@GuardedBy("evictionLock")
int evictFromWindow() {
int candidates = 0;
//查看window queue的頭部節(jié)點(diǎn)
Node<K, V> node = accessOrderWindowDeque().peek();
//如果window區(qū)超過(guò)了最大的限制,那么就要把“多出來(lái)”的記錄做處理
while (windowWeightedSize() > windowMaximum()) {
// The pending operations will adjust the size to reflect the correct weight
if (node == null) {
break;
}
//下一個(gè)節(jié)點(diǎn)
Node<K, V> next = node.getNextInAccessOrder();
if (node.getWeight() != 0) {
//把node定位在probation區(qū)
node.makeMainProbation();
//從window區(qū)去掉
accessOrderWindowDeque().remove(node);
//加入到probation queue,相當(dāng)于把節(jié)點(diǎn)移動(dòng)到probation區(qū)(晉升了)
accessOrderProbationDeque().add(node);
candidates++;
//因?yàn)橐瞥艘粋€(gè)節(jié)點(diǎn),所以需要調(diào)整window的size
setWindowWeightedSize(windowWeightedSize() - node.getPolicyWeight());
}
//處理下一個(gè)節(jié)點(diǎn)
node = next;
}
return candidates;
}
evictFromMain方法:
/**
* Evicts entries from the main space if the cache exceeds the maximum capacity. The main space
* determines whether admitting an entry (coming from the window space) is preferable to retaining
* the eviction policy's victim. This is decision is made using a frequency filter so that the
* least frequently used entry is removed.
*
* The window space candidates were previously placed in the MRU position and the eviction
* policy's victim is at the LRU position. The two ends of the queue are evaluated while an
* eviction is required. The number of remaining candidates is provided and decremented on
* eviction, so that when there are no more candidates the victim is evicted.
*
* @param candidates the number of candidate entries evicted from the window space
*/
//根據(jù)W-TinyLFU,從window晉升過(guò)來(lái)的要跟probation區(qū)的進(jìn)行“PK”,勝者才能留下
@GuardedBy("evictionLock")
void evictFromMain(int candidates) {
int victimQueue = PROBATION;
//victim是probation queue的頭部
Node<K, V> victim = accessOrderProbationDeque().peekFirst();
//candidate是probation queue的尾部,也就是剛從window晉升來(lái)的
Node<K, V> candidate = accessOrderProbationDeque().peekLast();
//當(dāng)cache不夠容量時(shí)才做處理
while (weightedSize() > maximum()) {
// Stop trying to evict candidates and always prefer the victim
if (candidates == 0) {
candidate = null;
}
//對(duì)candidate為null且victim為bull的處理
if ((candidate == null) && (victim == null)) {
if (victimQueue == PROBATION) {
victim = accessOrderProtectedDeque().peekFirst();
victimQueue = PROTECTED;
continue;
} else if (victimQueue == PROTECTED) {
victim = accessOrderWindowDeque().peekFirst();
victimQueue = WINDOW;
continue;
}
// The pending operations will adjust the size to reflect the correct weight
break;
}
//對(duì)節(jié)點(diǎn)的weight為0的處理
if ((victim != null) && (victim.getPolicyWeight() == 0)) {
victim = victim.getNextInAccessOrder();
continue;
} else if ((candidate != null) && (candidate.getPolicyWeight() == 0)) {
candidate = candidate.getPreviousInAccessOrder();
candidates--;
continue;
}
// Evict immediately if only one of the entries is present
if (victim == null) {
@SuppressWarnings("NullAway")
Node<K, V> previous = candidate.getPreviousInAccessOrder();
Node<K, V> evict = candidate;
candidate = previous;
candidates--;
evictEntry(evict, RemovalCause.SIZE, 0L);
continue;
} else if (candidate == null) {
Node<K, V> evict = victim;
victim = victim.getNextInAccessOrder();
evictEntry(evict, RemovalCause.SIZE, 0L);
continue;
}
// Evict immediately if an entry was collected
K victimKey = victim.getKey();
K candidateKey = candidate.getKey();
if (victimKey == null) {
@NonNull Node<K, V> evict = victim;
victim = victim.getNextInAccessOrder();
evictEntry(evict, RemovalCause.COLLECTED, 0L);
continue;
} else if (candidateKey == null) {
candidates--;
@NonNull Node<K, V> evict = candidate;
candidate = candidate.getPreviousInAccessOrder();
evictEntry(evict, RemovalCause.COLLECTED, 0L);
continue;
}
//放不下的節(jié)點(diǎn)直接處理掉
if (candidate.getPolicyWeight() > maximum()) {
candidates--;
Node<K, V> evict = candidate;
candidate = candidate.getPreviousInAccessOrder();
evictEntry(evict, RemovalCause.SIZE, 0L);
continue;
}
//根據(jù)節(jié)點(diǎn)的統(tǒng)計(jì)頻率frequency來(lái)做比較,看看要處理掉victim還是candidate
//admit是具體的比較規(guī)則,看下面
candidates--;
//如果candidate勝出則淘汰victim
if (admit(candidateKey, victimKey)) {
Node<K, V> evict = victim;
victim = victim.getNextInAccessOrder();
evictEntry(evict, RemovalCause.SIZE, 0L);
candidate = candidate.getPreviousInAccessOrder();
} else {
//如果是victim勝出,則淘汰candidate
Node<K, V> evict = candidate;
candidate = candidate.getPreviousInAccessOrder();
evictEntry(evict, RemovalCause.SIZE, 0L);
}
}
}
/**
* Determines if the candidate should be accepted into the main space, as determined by its
* frequency relative to the victim. A small amount of randomness is used to protect against hash
* collision attacks, where the victim's frequency is artificially raised so that no new entries
* are admitted.
*
* @param candidateKey the key for the entry being proposed for long term retention
* @param victimKey the key for the entry chosen by the eviction policy for replacement
* @return if the candidate should be admitted and the victim ejected
*/
@GuardedBy("evictionLock")
boolean admit(K candidateKey, K victimKey) {
//分別獲取victim和candidate的統(tǒng)計(jì)頻率
//frequency這個(gè)方法的原理和實(shí)現(xiàn)上面已經(jīng)解釋了
int victimFreq = frequencySketch().frequency(victimKey);
int candidateFreq = frequencySketch().frequency(candidateKey);
//誰(shuí)大誰(shuí)贏
if (candidateFreq > victimFreq) {
return true;
//如果相等,candidate小于5都當(dāng)輸了
} else if (candidateFreq <= 5) {
// The maximum frequency is 15 and halved to 7 after a reset to age the history. An attack
// exploits that a hot candidate is rejected in favor of a hot victim. The threshold of a warm
// candidate reduces the number of random acceptances to minimize the impact on the hit rate.
return false;
}
//如果相等且candidate大于5,則隨機(jī)淘汰一個(gè)
int random = ThreadLocalRandom.current().nextInt();
return ((random & 127) == 0);
}
climb方法主要是用來(lái)調(diào)整 window size 的,使得 Caffeine 可以適應(yīng)你的應(yīng)用類(lèi)型(如 OLAP 或 OLTP)表現(xiàn)出最佳的命中率。
下圖是官方測(cè)試的數(shù)據(jù):

我們看看 window size 的調(diào)整是怎么實(shí)現(xiàn)的。
調(diào)整時(shí)用到的默認(rèn)比例數(shù)據(jù):
//與上次命中率之差的閾值
static final double HILL_CLIMBER_RESTART_THRESHOLD = 0.05d;
//步長(zhǎng)(調(diào)整)的大小(跟最大值maximum的比例)
static final double HILL_CLIMBER_STEP_PERCENT = 0.0625d;
//步長(zhǎng)的衰減比例
static final double HILL_CLIMBER_STEP_DECAY_RATE = 0.98d;
/** Adapts the eviction policy to towards the optimal recency / frequency configuration. */
//climb方法的主要作用就是動(dòng)態(tài)調(diào)整window區(qū)的大小(相應(yīng)的,main區(qū)的大小也會(huì)發(fā)生變化,兩個(gè)之和為100%)。
//因?yàn)閰^(qū)域的大小發(fā)生了變化,那么區(qū)域內(nèi)的數(shù)據(jù)也可能需要發(fā)生相應(yīng)的移動(dòng)。
@GuardedBy("evictionLock")
void climb() {
if (!evicts()) {
return;
}
//確定window需要調(diào)整的大小
determineAdjustment();
//如果protected區(qū)有溢出,把溢出部分移動(dòng)到probation區(qū)。因?yàn)橄旅娴牟僮饔锌赡苄枰{(diào)整到protected區(qū)。
demoteFromMainProtected();
long amount = adjustment();
if (amount == 0) {
return;
} else if (amount > 0) {
//增加window的大小
increaseWindow();
} else {
//減少window的大小
decreaseWindow();
}
}
下面分別展開(kāi)每個(gè)方法來(lái)解釋?zhuān)?/p>
/** Calculates the amount to adapt the window by and sets {@link #adjustment()} accordingly. */
@GuardedBy("evictionLock")
void determineAdjustment() {
//如果frequencySketch還沒(méi)初始化,則返回
if (frequencySketch().isNotInitialized()) {
setPreviousSampleHitRate(0.0);
setMissesInSample(0);
setHitsInSample(0);
return;
}
//總請(qǐng)求量 = 命中 + miss
int requestCount = hitsInSample() + missesInSample();
//沒(méi)達(dá)到sampleSize則返回
//默認(rèn)下sampleSize = 10 * maximum。用sampleSize來(lái)判斷緩存是否足夠”熱“。
if (requestCount < frequencySketch().sampleSize) {
return;
}
//命中率的公式 = 命中 / 總請(qǐng)求
double hitRate = (double) hitsInSample() / requestCount;
//命中率的差值
double hitRateChange = hitRate - previousSampleHitRate();
//本次調(diào)整的大小,是由命中率的差值和上次的stepSize決定的
double amount = (hitRateChange >= 0) ? stepSize() : -stepSize();
//下次的調(diào)整大小:如果命中率的之差大于0.05,則重置為0.065 * maximum,否則按照0.98來(lái)進(jìn)行衰減
double nextStepSize = (Math.abs(hitRateChange) >= HILL_CLIMBER_RESTART_THRESHOLD)
? HILL_CLIMBER_STEP_PERCENT * maximum() * (amount >= 0 ? 1 : -1)
: HILL_CLIMBER_STEP_DECAY_RATE * amount;
setPreviousSampleHitRate(hitRate);
setAdjustment((long) amount);
setStepSize(nextStepSize);
setMissesInSample(0);
setHitsInSample(0);
}
/** Transfers the nodes from the protected to the probation region if it exceeds the maximum. */
//這個(gè)方法比較簡(jiǎn)單,減少protected區(qū)溢出的部分
@GuardedBy("evictionLock")
void demoteFromMainProtected() {
long mainProtectedMaximum = mainProtectedMaximum();
long mainProtectedWeightedSize = mainProtectedWeightedSize();
if (mainProtectedWeightedSize <= mainProtectedMaximum) {
return;
}
for (int i = 0; i < QUEUE_TRANSFER_THRESHOLD; i++) {
if (mainProtectedWeightedSize <= mainProtectedMaximum) {
break;
}
Node<K, V> demoted = accessOrderProtectedDeque().poll();
if (demoted == null) {
break;
}
demoted.makeMainProbation();
accessOrderProbationDeque().add(demoted);
mainProtectedWeightedSize -= demoted.getPolicyWeight();
}
setMainProtectedWeightedSize(mainProtectedWeightedSize);
}
/**
* Increases the size of the admission window by shrinking the portion allocated to the main
* space. As the main space is partitioned into probation and protected regions (80% / 20%), for
* simplicity only the protected is reduced. If the regions exceed their maximums, this may cause
* protected items to be demoted to the probation region and probation items to be demoted to the
* admission window.
*/
//增加window區(qū)的大小,這個(gè)方法比較簡(jiǎn)單,思路就像我上面說(shuō)的
@GuardedBy("evictionLock")
void increaseWindow() {
if (mainProtectedMaximum() == 0) {
return;
}
long quota = Math.min(adjustment(), mainProtectedMaximum());
setMainProtectedMaximum(mainProtectedMaximum() - quota);
setWindowMaximum(windowMaximum() + quota);
demoteFromMainProtected();
for (int i = 0; i < QUEUE_TRANSFER_THRESHOLD; i++) {
Node<K, V> candidate = accessOrderProbationDeque().peek();
boolean probation = true;
if ((candidate == null) || (quota < candidate.getPolicyWeight())) {
candidate = accessOrderProtectedDeque().peek();
probation = false;
}
if (candidate == null) {
break;
}
int weight = candidate.getPolicyWeight();
if (quota < weight) {
break;
}
quota -= weight;
if (probation) {
accessOrderProbationDeque().remove(candidate);
} else {
setMainProtectedWeightedSize(mainProtectedWeightedSize() - weight);
accessOrderProtectedDeque().remove(candidate);
}
setWindowWeightedSize(windowWeightedSize() + weight);
accessOrderWindowDeque().add(candidate);
candidate.makeWindow();
}
setMainProtectedMaximum(mainProtectedMaximum() + quota);
setWindowMaximum(windowMaximum() - quota);
setAdjustment(quota);
}
/** Decreases the size of the admission window and increases the main's protected region. */
//同上increaseWindow差不多,反操作
@GuardedBy("evictionLock")
void decreaseWindow() {
if (windowMaximum() <= 1) {
return;
}
long quota = Math.min(-adjustment(), Math.max(0, windowMaximum() - 1));
setMainProtectedMaximum(mainProtectedMaximum() + quota);
setWindowMaximum(windowMaximum() - quota);
for (int i = 0; i < QUEUE_TRANSFER_THRESHOLD; i++) {
Node<K, V> candidate = accessOrderWindowDeque().peek();
if (candidate == null) {
break;
}
int weight = candidate.getPolicyWeight();
if (quota < weight) {
break;
}
quota -= weight;
setMainProtectedWeightedSize(mainProtectedWeightedSize() + weight);
setWindowWeightedSize(windowWeightedSize() - weight);
accessOrderWindowDeque().remove(candidate);
accessOrderProbationDeque().add(candidate);
candidate.makeMainProbation();
}
setMainProtectedMaximum(mainProtectedMaximum() - quota);
setWindowMaximum(windowMaximum() + quota);
setAdjustment(-quota);
}
以上,是 Caffeine 的 W-TinyLFU 策略的設(shè)計(jì)原理及代碼實(shí)現(xiàn)解析。
異步的高性能讀寫(xiě)
一般的緩存每次對(duì)數(shù)據(jù)處理完之后(讀的話,已經(jīng)存在則直接返回,不存在則 load 數(shù)據(jù),保存,再返回;寫(xiě)的話,則直接插入或更新),但是因?yàn)橐S護(hù)一些淘汰策略,則需要一些額外的操作,諸如:
計(jì)算和比較數(shù)據(jù)的是否過(guò)期 統(tǒng)計(jì)頻率(像 LFU 或其變種) 維護(hù) read queue 和 write queue 淘汰符合條件的數(shù)據(jù) 等等。。。
這種數(shù)據(jù)的讀寫(xiě)伴隨著緩存狀態(tài)的變更,Guava Cache 的做法是把這些操作和讀寫(xiě)操作放在一起,在一個(gè)同步加鎖的操作中完成,雖然 Guava Cache 巧妙地利用了 JDK 的 ConcurrentHashMap(分段鎖或者無(wú)鎖 CAS)來(lái)降低鎖的密度,達(dá)到提高并發(fā)度的目的。但是,對(duì)于一些熱點(diǎn)數(shù)據(jù),這種做法還是避免不了頻繁的鎖競(jìng)爭(zhēng)。Caffeine 借鑒了數(shù)據(jù)庫(kù)系統(tǒng)的 WAL(Write-Ahead Logging)思想,即先寫(xiě)日志再執(zhí)行操作,這種思想同樣適合緩存的,執(zhí)行讀寫(xiě)操作時(shí),先把操作記錄在緩沖區(qū),然后在合適的時(shí)機(jī)異步、批量地執(zhí)行緩沖區(qū)中的內(nèi)容。但在執(zhí)行緩沖區(qū)的內(nèi)容時(shí),也是需要在緩沖區(qū)加上同步鎖的,不然存在并發(fā)問(wèn)題,只不過(guò)這樣就可以把對(duì)鎖的競(jìng)爭(zhēng)從緩存數(shù)據(jù)轉(zhuǎn)移到對(duì)緩沖區(qū)上。
ReadBuffer
在 Caffeine 的內(nèi)部實(shí)現(xiàn)中,為了很好的支持不同的 Features(如 Eviction,Removal,Refresh,Statistics,Cleanup,Policy 等等),擴(kuò)展了很多子類(lèi),它們共同的父類(lèi)是BoundedLocalCache,而readBuffer就是作為它們共有的屬性,即都是用一樣的 readBuffer,看定義:
final Buffer<Node<K, V>> readBuffer;
readBuffer = evicts() || collectKeys() || collectValues() || expiresAfterAccess()
? new BoundedBuffer<>()
: Buffer.disabled();
上面提到 Caffeine 對(duì)每次緩存的讀操作都會(huì)觸發(fā)afterRead
/**
* Performs the post-processing work required after a read.
*
* @param node the entry in the page replacement policy
* @param now the current time, in nanoseconds
* @param recordHit if the hit count should be incremented
*/
void afterRead(Node<K, V> node, long now, boolean recordHit) {
if (recordHit) {
statsCounter().recordHits(1);
}
//把記錄加入到readBuffer
//判斷是否需要立即處理readBuffer
//注意這里無(wú)論offer是否成功都可以走下去的,即允許寫(xiě)入readBuffer丟失,因?yàn)檫@個(gè)
boolean delayable = skipReadBuffer() || (readBuffer.offer(node) != Buffer.FULL);
if (shouldDrainBuffers(delayable)) {
scheduleDrainBuffers();
}
refreshIfNeeded(node, now);
}
/**
* Returns whether maintenance work is needed.
*
* @param delayable if draining the read buffer can be delayed
*/
//caffeine用了一組狀態(tài)來(lái)定義和管理“維護(hù)”的過(guò)程
boolean shouldDrainBuffers(boolean delayable) {
switch (drainStatus()) {
case IDLE:
return !delayable;
case REQUIRED:
return true;
case PROCESSING_TO_IDLE:
case PROCESSING_TO_REQUIRED:
return false;
default:
throw new IllegalStateException();
}
}
重點(diǎn)看BoundedBuffer
/**
* A striped, non-blocking, bounded buffer.
*
* @author [email protected] (Ben Manes)
* @param <E> the type of elements maintained by this buffer
*/
final class BoundedBuffer<E> extends StripedBuffer<E>
它是一個(gè) striped、非阻塞、有界限的 buffer,繼承于StripedBuffer類(lèi)。下面看看StripedBuffer的實(shí)現(xiàn):
/**
* A base class providing the mechanics for supporting dynamic striping of bounded buffers. This
* implementation is an adaption of the numeric 64-bit {@link java.util.concurrent.atomic.Striped64}
* class, which is used by atomic counters. The approach was modified to lazily grow an array of
* buffers in order to minimize memory usage for caches that are not heavily contended on.
*
* @author [email protected] (Doug Lea)
* @author [email protected] (Ben Manes)
*/
abstract class StripedBuffer<E> implements Buffer<E>
這個(gè)StripedBuffer設(shè)計(jì)的思想是跟Striped64類(lèi)似的,通過(guò)擴(kuò)展結(jié)構(gòu)把競(jìng)爭(zhēng)熱點(diǎn)分離。
具體實(shí)現(xiàn)是這樣的,StripedBuffer維護(hù)一個(gè)Buffer[]數(shù)組,每個(gè)元素就是一個(gè)RingBuffer,每個(gè)線程用自己threadLocalRandomProbe屬性作為 hash 值,這樣就相當(dāng)于每個(gè)線程都有自己“專(zhuān)屬”的RingBuffer,就不會(huì)產(chǎn)生競(jìng)爭(zhēng)啦,而不是用 key 的hashCode作為 hash 值,因?yàn)闀?huì)產(chǎn)生熱點(diǎn)數(shù)據(jù)問(wèn)題。
看看StripedBuffer的屬性
/** Table of buffers. When non-null, size is a power of 2. */
//RingBuffer數(shù)組
transient volatile Buffer<E> @Nullable[] table;
//當(dāng)進(jìn)行resize時(shí),需要整個(gè)table鎖住。tableBusy作為CAS的標(biāo)記。
static final long TABLE_BUSY = UnsafeAccess.objectFieldOffset(StripedBuffer.class, "tableBusy");
static final long PROBE = UnsafeAccess.objectFieldOffset(Thread.class, "threadLocalRandomProbe");
/** Number of CPUS. */
static final int NCPU = Runtime.getRuntime().availableProcessors();
/** The bound on the table size. */
//table最大size
static final int MAXIMUM_TABLE_SIZE = 4 * ceilingNextPowerOfTwo(NCPU);
/** The maximum number of attempts when trying to expand the table. */
//如果發(fā)生競(jìng)爭(zhēng)時(shí)(CAS失敗)的嘗試次數(shù)
static final int ATTEMPTS = 3;
/** Table of buffers. When non-null, size is a power of 2. */
//核心數(shù)據(jù)結(jié)構(gòu)
transient volatile Buffer<E> @Nullable[] table;
/** Spinlock (locked via CAS) used when resizing and/or creating Buffers. */
transient volatile int tableBusy;
/** CASes the tableBusy field from 0 to 1 to acquire lock. */
final boolean casTableBusy() {
return UnsafeAccess.UNSAFE.compareAndSwapInt(this, TABLE_BUSY, 0, 1);
}
/**
* Returns the probe value for the current thread. Duplicated from ThreadLocalRandom because of
* packaging restrictions.
*/
static final int getProbe() {
return UnsafeAccess.UNSAFE.getInt(Thread.currentThread(), PROBE);
}
offer方法,當(dāng)沒(méi)初始化或存在競(jìng)爭(zhēng)時(shí),則擴(kuò)容為 2 倍。
實(shí)際是調(diào)用RingBuffer的 offer 方法,把數(shù)據(jù)追加到RingBuffer后面。
@Override
public int offer(E e) {
int mask;
int result = 0;
Buffer<E> buffer;
//是否不存在競(jìng)爭(zhēng)
boolean uncontended = true;
Buffer<E>[] buffers = table
//是否已經(jīng)初始化
if ((buffers == null)
|| (mask = buffers.length - 1) < 0
//用thread的隨機(jī)值作為hash值,得到對(duì)應(yīng)位置的RingBuffer
|| (buffer = buffers[getProbe() & mask]) == null
//檢查追加到RingBuffer是否成功
|| !(uncontended = ((result = buffer.offer(e)) != Buffer.FAILED))) {
//其中一個(gè)符合條件則進(jìn)行擴(kuò)容
expandOrRetry(e, uncontended);
}
return result;
}
/**
* Handles cases of updates involving initialization, resizing, creating new Buffers, and/or
* contention. See above for explanation. This method suffers the usual non-modularity problems of
* optimistic retry code, relying on rechecked sets of reads.
*
* @param e the element to add
* @param wasUncontended false if CAS failed before call
*/
//這個(gè)方法比較長(zhǎng),但思路還是相對(duì)清晰的。
@SuppressWarnings("PMD.ConfusingTernary")
final void expandOrRetry(E e, boolean wasUncontended) {
int h;
if ((h = getProbe()) == 0) {
ThreadLocalRandom.current(); // force initialization
h = getProbe();
wasUncontended = true;
}
boolean collide = false; // True if last slot nonempty
for (int attempt = 0; attempt < ATTEMPTS; attempt++) {
Buffer<E>[] buffers;
Buffer<E> buffer;
int n;
if (((buffers = table) != null) && ((n = buffers.length) > 0)) {
if ((buffer = buffers[(n - 1) & h]) == null) {
if ((tableBusy == 0) && casTableBusy()) { // Try to attach new Buffer
boolean created = false;
try { // Recheck under lock
Buffer<E>[] rs;
int mask, j;
if (((rs = table) != null) && ((mask = rs.length) > 0)
&& (rs[j = (mask - 1) & h] == null)) {
rs[j] = create(e);
created = true;
}
} finally {
tableBusy = 0;
}
if (created) {
break;
}
continue; // Slot is now non-empty
}
collide = false;
} else if (!wasUncontended) { // CAS already known to fail
wasUncontended = true; // Continue after rehash
} else if (buffer.offer(e) != Buffer.FAILED) {
break;
} else if (n >= MAXIMUM_TABLE_SIZE || table != buffers) {
collide = false; // At max size or stale
} else if (!collide) {
collide = true;
} else if (tableBusy == 0 && casTableBusy()) {
try {
if (table == buffers) { // Expand table unless stale
table = Arrays.copyOf(buffers, n << 1);
}
} finally {
tableBusy = 0;
}
collide = false;
continue; // Retry with expanded table
}
h = advanceProbe(h);
} else if ((tableBusy == 0) && (table == buffers) && casTableBusy()) {
boolean init = false;
try { // Initialize table
if (table == buffers) {
@SuppressWarnings({"unchecked", "rawtypes"})
Buffer<E>[] rs = new Buffer[1];
rs[0] = create(e);
table = rs;
init = true;
}
} finally {
tableBusy = 0;
}
if (init) {
break;
}
}
}
}
最后看看RingBuffer,注意RingBuffer是BoundedBuffer的內(nèi)部類(lèi)。
/** The maximum number of elements per buffer. */
static final int BUFFER_SIZE = 16;
// Assume 4-byte references and 64-byte cache line (16 elements per line)
//256長(zhǎng)度,但是是以16為單位,所以最多存放16個(gè)元素
static final int SPACED_SIZE = BUFFER_SIZE << 4;
static final int SPACED_MASK = SPACED_SIZE - 1;
static final int OFFSET = 16;
//RingBuffer數(shù)組
final AtomicReferenceArray<E> buffer;
//插入方法
@Override
public int offer(E e) {
long head = readCounter;
long tail = relaxedWriteCounter();
//用head和tail來(lái)限制個(gè)數(shù)
long size = (tail - head);
if (size >= SPACED_SIZE) {
return Buffer.FULL;
}
//tail追加16
if (casWriteCounter(tail, tail + OFFSET)) {
//用tail“取余”得到下標(biāo)
int index = (int) (tail & SPACED_MASK);
//用unsafe.putOrderedObject設(shè)值
buffer.lazySet(index, e);
return Buffer.SUCCESS;
}
//如果CAS失敗則返回失敗
return Buffer.FAILED;
}
//用consumer來(lái)處理buffer的數(shù)據(jù)
@Override
public void drainTo(Consumer<E> consumer) {
long head = readCounter;
long tail = relaxedWriteCounter();
//判斷數(shù)據(jù)多少
long size = (tail - head);
if (size == 0) {
return;
}
do {
int index = (int) (head & SPACED_MASK);
E e = buffer.get(index);
if (e == null) {
// not published yet
break;
}
buffer.lazySet(index, null);
consumer.accept(e);
//head也跟tail一樣,每次遞增16
head += OFFSET;
} while (head != tail);
lazySetReadCounter(head);
}
注意,ring buffer 的 size(固定是 16 個(gè))是不變的,變的是 head 和 tail 而已。
總的來(lái)說(shuō)ReadBuffer有如下特點(diǎn):
使用 Striped-RingBuffer來(lái)提升對(duì) buffer 的讀寫(xiě)用 thread 的 hash 來(lái)避開(kāi)熱點(diǎn) key 的競(jìng)爭(zhēng) 允許寫(xiě)入的丟失
WriteBuffer
writeBuffer跟readBuffer不一樣,主要體現(xiàn)在使用場(chǎng)景的不一樣。本來(lái)緩存的一般場(chǎng)景是讀多寫(xiě)少的,讀的并發(fā)會(huì)更高,且 afterRead 顯得沒(méi)那么重要,允許延遲甚至丟失。寫(xiě)不一樣,寫(xiě)afterWrite不允許丟失,且要求盡量馬上執(zhí)行。Caffeine 使用MPSC(Multiple Producer / Single Consumer)作為 buffer 數(shù)組,實(shí)現(xiàn)在MpscGrowableArrayQueue類(lèi),它是仿照JCTools的MpscGrowableArrayQueue來(lái)寫(xiě)的。
MPSC 允許無(wú)鎖的高并發(fā)寫(xiě)入,但只允許一個(gè)消費(fèi)者,同時(shí)也犧牲了部分操作。
MPSC 我打算另外分析,這里不展開(kāi)了。
TimerWheel
除了支持expireAfterAccess和expireAfterWrite之外(Guava Cache 也支持這兩個(gè)特性),Caffeine 還支持expireAfter。因?yàn)?code style="font-size: 14px;padding: 2px 4px;border-radius: 4px;margin-right: 2px;margin-left: 2px;background-color: rgba(27, 31, 35, 0.05);font-family: 'Operator Mono', Consolas, Monaco, Menlo, monospace;word-break: break-all;color: rgb(150, 84, 181);">expireAfterAccess和expireAfterWrite都只能是固定的過(guò)期時(shí)間,這可能滿(mǎn)足不了某些場(chǎng)景,譬如記錄的過(guò)期時(shí)間是需要根據(jù)某些條件而不一樣的,這就需要用戶(hù)自定義過(guò)期時(shí)間。
先看看expireAfter的用法
private static LoadingCache<String, String> cache = Caffeine.newBuilder()
.maximumSize(256L)
.initialCapacity(1)
//.expireAfterAccess(2, TimeUnit.DAYS)
//.expireAfterWrite(2, TimeUnit.HOURS)
.refreshAfterWrite(1, TimeUnit.HOURS)
//自定義過(guò)期時(shí)間
.expireAfter(new Expiry<String, String>() {
//返回創(chuàng)建后的過(guò)期時(shí)間
@Override
public long expireAfterCreate(@NonNull String key, @NonNull String value, long currentTime) {
return 0;
}
//返回更新后的過(guò)期時(shí)間
@Override
public long expireAfterUpdate(@NonNull String key, @NonNull String value, long currentTime, @NonNegative long currentDuration) {
return 0;
}
//返回讀取后的過(guò)期時(shí)間
@Override
public long expireAfterRead(@NonNull String key, @NonNull String value, long currentTime, @NonNegative long currentDuration) {
return 0;
}
})
.recordStats()
.build(new CacheLoader<String, String>() {
@Nullable
@Override
public String load(@NonNull String key) throws Exception {
return "value_" + key;
}
});
通過(guò)自定義過(guò)期時(shí)間,使得不同的 key 可以動(dòng)態(tài)的得到不同的過(guò)期時(shí)間。
注意,我把expireAfterAccess和expireAfterWrite注釋了,因?yàn)檫@兩個(gè)特性不能跟expireAfter一起使用。
而當(dāng)使用了expireAfter特性后,Caffeine 會(huì)啟用一種叫“時(shí)間輪”的算法來(lái)實(shí)現(xiàn)這個(gè)功能。更多關(guān)于時(shí)間輪的介紹,可以看我的文章HashedWheelTimer 時(shí)間輪原理分析[6]。
好,重點(diǎn)來(lái)了,為什么要用時(shí)間輪?
對(duì)expireAfterAccess和expireAfterWrite的實(shí)現(xiàn)是用一個(gè)AccessOrderDeque雙端隊(duì)列,它是 FIFO 的,因?yàn)樗鼈兊倪^(guò)期時(shí)間是固定的,所以在隊(duì)列頭的數(shù)據(jù)肯定是最早過(guò)期的,要處理過(guò)期數(shù)據(jù)時(shí),只需要首先看看頭部是否過(guò)期,然后再挨個(gè)檢查就可以了。但是,如果過(guò)期時(shí)間不一樣的話,這需要對(duì)accessOrderQueue進(jìn)行排序&插入,這個(gè)代價(jià)太大了。于是,Caffeine 用了一種更加高效、優(yōu)雅的算法-時(shí)間輪。
時(shí)間輪的結(jié)構(gòu):

因?yàn)樵谖业膶?duì)時(shí)間輪分析的文章里已經(jīng)說(shuō)了時(shí)間輪的原理和機(jī)制了,所以我就不展開(kāi) Caffeine 對(duì)時(shí)間輪的實(shí)現(xiàn)了。
Caffeine 對(duì)時(shí)間輪的實(shí)現(xiàn)在TimerWheel,它是一種多層時(shí)間輪(hierarchical timing wheels )。
看看元素加入到時(shí)間輪的schedule方法:
/**
* Schedules a timer event for the node.
*
* @param node the entry in the cache
*/
public void schedule(@NonNull Node<K, V> node) {
Node<K, V> sentinel = findBucket(node.getVariableTime());
link(sentinel, node);
}
/**
* Determines the bucket that the timer event should be added to.
*
* @param time the time when the event fires
* @return the sentinel at the head of the bucket
*/
Node<K, V> findBucket(long time) {
long duration = time - nanos;
int length = wheel.length - 1;
for (int i = 0; i < length; i++) {
if (duration < SPANS[i + 1]) {
long ticks = (time >>> SHIFT[i]);
int index = (int) (ticks & (wheel[i].length - 1));
return wheel[i][index];
}
}
return wheel[length][0];
}
/** Adds the entry at the tail of the bucket's list. */
void link(Node<K, V> sentinel, Node<K, V> node) {
node.setPreviousInVariableOrder(sentinel.getPreviousInVariableOrder());
node.setNextInVariableOrder(sentinel);
sentinel.getPreviousInVariableOrder().setNextInVariableOrder(node);
sentinel.setPreviousInVariableOrder(node);
}
其他
Caffeine 還有其他的優(yōu)化性能的手段,如使用軟引用和弱引用、消除偽共享、CompletableFuture異步等等。
總結(jié)
Caffeien 是一個(gè)優(yōu)秀的本地緩存,通過(guò)使用 W-TinyLFU 算法, 高性能的 readBuffer 和 WriteBuffer,時(shí)間輪算法等,使得它擁有高性能,高命中率(near optimal),低內(nèi)存占用等特點(diǎn)。
參考資料
TinyLFU 論文[7]
Design Of A Modern Cache[8]
Design Of A Modern Cache—Part Deux[9]
Caffeine 的 github[10]
參考資料
Caffeine: https://github.com/ben-manes/caffeine
[2]這里: https://albenw.github.io/posts/df42dc84/
[3]Benchmarks: https://github.com/ben-manes/caffeine/wiki/Benchmarks
[4]官方API說(shuō)明文檔: https://github.com/ben-manes/caffeine/wiki
[5]這里: https://github.com/ben-manes/caffeine/wiki/Guava
[6]HashedWheelTimer時(shí)間輪原理分析: https://albenw.github.io/posts/ec8df8c/
[7]TinyLFU論文: https://arxiv.org/abs/1512.00727
[8]Design Of A Modern Cache: http://highscalability.com/blog/2016/1/25/design-of-a-modern-cache.html
[9]Design Of A Modern Cache—Part Deux: http://highscalability.com/blog/2019/2/25/design-of-a-modern-cachepart-deux.html
[10]Caffeine的github: https://github.com/ben-manes/caffeine
來(lái)源:albenw.github.io/posts/a4ae1aa2
覺(jué)得不錯(cuò),請(qǐng)點(diǎn)個(gè)在看呀
