README
¶
GigaCache
GigaCache 是一个 Golang 编写的高性能缓存库，使用紧凑的字节数组作为数据容器，为 GB 级序列化数据而设计，支持设置过期时间与淘汰机制，相比 stdmap 有更快的速度，更高的内存效率，和更小的 GC 延迟。
特性：
- 高效管理序列化的数据，存储相同数据量相比
map[string]string内存占用减少 60%，性能提升 40%+，GC 延迟降低 30%。 - 采用分片技术减小锁粒度，并分块管理数据
- 键值对独立的过期时间支持，使用定期淘汰策略驱逐过期的键值对
- 内置迁移算法，定期整理碎片空间，以释放内存
- 类似于
bigcache规避 GC 的设计，上亿数据集的 P99 延迟在微秒级别
你可以阅读 博客文档 了解更多的技术细节。
性能
下面是插入 2000 万条数据的性能对比测试，GigaCache 的插入速度相比 stdmap 提升 40%+，内存占用减少了 60% 左右，GC 延迟降低了 30%。
gigacache
entries: 20000000
alloc: 1103 mb
gcsys: 7 mb
heap inuse: 1103 mb
heap object: 2598 k
gc: 13
pause: 2.092264ms
cost: 13.172292035s
stdmap
entries: 20000000
alloc: 2703 mb
gcsys: 16 mb
heap inuse: 2709 mb
heap object: 29597 k
gc: 11
pause: 2.702983ms
cost: 18.730218048s
详细测试
运行 make bench 命令进行详细测试:
goos: linux
goarch: amd64
pkg: github.com/xgzlucario/GigaCache
cpu: 13th Gen Intel(R) Core(TM) i5-13600KF
BenchmarkSet/stdmap-20 3978168 360.2 ns/op 225 B/op 3 allocs/op
BenchmarkSet/cache-20 5956052 234.2 ns/op 95 B/op 2 allocs/op
BenchmarkSet/cache/disableEvict-20 7046114 214.1 ns/op 104 B/op 2 allocs/op
BenchmarkGet/stdmap-20 8400390 152.7 ns/op 15 B/op 1 allocs/op
BenchmarkGet/cache-20 10187661 117.5 ns/op 16 B/op 2 allocs/op
BenchmarkScan/stdmap-20 151 7814006 ns/op 0 B/op 0 allocs/op
BenchmarkScan/cache-20 100 11475189 ns/op 0 B/op 0 allocs/op
BenchmarkRemove/stdmap-20 16358985 66.26 ns/op 16 B/op 1 allocs/op
BenchmarkRemove/cache-20 7386679 151.4 ns/op 16 B/op 2 allocs/op
BenchmarkMigrate/stdmap-20 676 1739740 ns/op 6441224 B/op 1685 allocs/op
BenchmarkMigrate/cache-20 343 3561328 ns/op 24620 B/op 1024 allocs/op
BenchmarkMigrate/cache/parallel-20 1183 1033018 ns/op 42423 B/op 2091 allocs/op
BenchmarkIdx/newIdx-20 1000000000 0.7566 ns/op 0 B/op 0 allocs/op
BenchmarkIdx/newIdxx-20 1000000000 0.3761 ns/op 0 B/op 0 allocs/op
PASS
测试环境
goos: linux
goarch: amd64
pkg: github.com/xgzlucario/GigaCache
cpu: AMD Ryzen 7 5800H with Radeon Graphics
使用
首先安装 GigaCache 到本地：
go get github.com/xgzlucario/GigaCache
运行下面的代码示例：
package main
import (
"fmt"
cache "github.com/xgzlucario/GigaCache"
)
func main() {
m := cache.New(cache.DefaultOptions)
m.Set("foo", []byte("bar"))
// Set with expired time.
m.SetEx("foo1", []byte("bar1"), time.Minute)
// Set with deadline.
m.SetTx("foo2", []byte("bar2"), time.Now().Add(time.Minute).UnixNano())
val, ts, ok := m.Get("foo")
fmt.Println(string(val), ok) // bar, (nanosecs), true
ok := m.Delete("foo1") // true
if ok {
// ...
}
// or Range cache
m.Scan(func(key, val []byte, ts int64) bool {
// ...
return false
})
}
内部架构
GigaCache
Key & Idx
Bucket
Documentation
¶
Index ¶
- Constants
- Variables
- func FastRand() uint32
- func FastRand64() uint64
- func GetNanoSec() int64
- func GetSec() uint32
- func MemHash(str string) uint64
- func SizeUvarint(x uint64) int
- type GigaCache
- func (c *GigaCache) Get(keyStr string) ([]byte, int64, bool)
- func (c *GigaCache) GetStats() (stats Stats)
- func (c *GigaCache) Migrate()
- func (c *GigaCache) Remove(keyStr string) bool
- func (c *GigaCache) Scan(callback Walker)
- func (c *GigaCache) Set(keyStr string, value []byte) bool
- func (c *GigaCache) SetEx(keyStr string, value []byte, duration time.Duration) bool
- func (c *GigaCache) SetTTL(keyStr string, expiration int64) bool
- func (c *GigaCache) SetTx(keyStr string, value []byte, expiration int64) bool
- type HashFn
- type Idx
- type Key
- type Options
- type Stats
- type Walker
Constants ¶
View Source
const (
KB = 1024
)
Variables ¶
Functions ¶
func FastRand64 ¶
func FastRand64() uint64
func MemHash ¶
MemHash is the hash function used by go map, it utilizes available hardware instructions (behaves as aes hash if aes instruction is available). NOTE: The hash seed changes for every process. So, this cannot be used as a persistent hash.
Types ¶
type GigaCache ¶
type GigaCache struct {
// contains filtered or unexported fields
}
GigaCache implements a key-value cache.
func (*GigaCache) Migrate ¶
func (c *GigaCache) Migrate()
Migrate transfers all data to new buckets.
func (*GigaCache) Scan ¶
Scan iterates over all alive key-value pairs without copying the data. DO NOT MODIFY the bytes as they are not copied.
type Options ¶
type Options struct {
// ShardCount is shard numbers of cache.
ShardCount uint32
// Default size of the bucket initial.
IndexSize int
BufferSize int
// EvictInterval indicates the frequency of execution of the eliminate algorithm.
// the higher the frequency, the more expired key-value pairs will be evicted,
// but accordingly it will slow down the overall performance,
// because the system needs to spend more time on probing and evicting.
EvictInterval uint8
// DisableEvict
// Set `true` when you don't need any expiration times.
DisableEvict bool
// Migrate threshold for a bucket to trigger a migration.
MigrateRatio float64
// HashFn is custom hash function, default is runtime.memhash.
HashFn HashFn
// ConcurrencySafe specifies whether RWLocker are required for multithreading safety.
ConcurrencySafe bool
}
Options is the configuration of GigaCache.
type Stats ¶
type Stats struct {
Len int
Conflicts int
Alloc uint64
Unused uint64
Migrates uint64
Evictions uint64
Probes uint64
}
Stats represents the runtime statistics of GigaCache.
func (Stats) EvictionRate ¶
EvictionRate calculates the percentage of evictions relative to probes.
func (Stats) UnusedRate ¶
UnusedRate calculates the percentage of unused space in the cache.
Source Files
Directories
Click to show internal directories.
Click to hide internal directories.