内存开销

内存开销测试通过Java Agent for Memory Measurements来计算运行时内存。这个大小可能受到引用压缩，对象填充等JVM因素影响。这个基准测试可以通过gradle memoryOverhead来执行内存基准测试。

通过在Java 5的[ConcurrentHashMap][3]上进行分支开发，Guava在每个缓存元素的所消耗的内存进行了优化。当使用软引用和弱饮用的时候这点优势会更加明显。因为Caffeine 是基于Java 8的[ConcurrentHashMap][3]进行封装的，需要额外的字段去移除一个被回收的字段。

Caffeine将会基于用法懒加载或是动态调整器内部的数据结构大小。这减少了内存开销并可以在使用量增加的时候(比如吞吐量)方便追加内存占用。

无界

Cache	Baseline	Entry
Caffeine	264 bytes	54 bytes (56 aligned)
Guava	1,032 bytes	58 bytes (64 aligned)

容量有界

Cache	Baseline	Entry
Caffeine	1,112 bytes	74 bytes (80 aligned)
Guava	1,384 bytes	74 bytes (80 aligned)

容量有界 & 根据访问时间过期

Cache	Baseline	Entry
Caffeine	1,152 bytes	82 bytes (88 aligned)
Guava	1,384 bytes	74 bytes (80 aligned)

容量有界 & 根据写入时间过期

Cache	Baseline	Entry
Caffeine	1,176 bytes	90 bytes (96 aligned)
Guava	1,544 bytes	90 bytes (96 aligned)

容量有界 & 根据写入时间刷新

Cache	Baseline	Entry
Caffeine	1,168 bytes	90 bytes (96 aligned)
Guava	27,200 bytes (?)	90 bytes (96 aligned)

权重有界

Cache	Baseline	Entry
Caffeine	1,120 bytes	82 bytes (88 aligned)
Guava	1,376 bytes	74 bytes (80 aligned)

根据访问时间过期

Cache	Baseline	Entry
Caffeine	1,008 bytes	82 bytes (88 aligned)
Guava	1,384 bytes	74 bytes (80 aligned)

根据写入时间过期

Cache	Baseline	Entry
Caffeine	1,008 bytes	82 bytes (88 aligned)
Guava	1,192 bytes	74 bytes (80 aligned)

根据访问时间 & 写入时间过期

Cache	Baseline	Entry
Caffeine	1,048 bytes	98 bytes (104 aligned)
Guava	1,544 bytes	90 bytes (96 aligned)

弱引用Keys

Cache	Baseline	Entry
Caffeine	560 bytes	98 bytes (104 aligned)
Guava	1,240 bytes	66 bytes (72 aligned)

弱引用Values

Cache	Baseline	Entry
Caffeine	560 bytes	98 bytes (104 aligned)
Guava	1,240 bytes	74 bytes (80 aligned)

弱引用Keys & Values

Cache	Baseline	Entry
Caffeine	616 bytes	130 bytes (136 aligned)
Guava	1,416 bytes	82 bytes (88 aligned)

弱引用 Keys & 软引用 Values

Cache	Baseline	Entry
Caffeine	616 bytes	146 bytes (152 aligned)
Guava	1,416 bytes	98 bytes (104 aligned)

软引用Values

Cache	Baseline	Entry
Caffeine	560 bytes	114 bytes (120 aligned)
Guava	1,240 bytes	90 bytes (96 aligned)

[3]: https://docs.oracle.com/javase/8/docs/api/java/util/concurrent/ConcurrentHashMap.html