New cache implementation and prefetch handing in mw/cache (#731)

* cache: add sharded cache implementation

Add Cache impl and a few tests. This cache is 256-way sharded, mainly
so each shard has it's own lock. The main cache structure is a readonly
jump plane into the right shard.

This should remove the single lock contention on the main lock and
provide more concurrent throughput - Obviously this hasn't been tested
or measured.

The key into the cache was made a uint32 (hash.fnv) and the hashing op
is not using strings.ToLower anymore remove any GC in that code path.

* here too

* Minimum shard size

* typos

* blurp

* small cleanups no defer

* typo

* Add freq based on Johns idea

* cherry-pick conflict resolv

* typo

* update from early code review from john

* add prefetch to the cache

* mw/cache: add prefetch

* remove println

* remove comment

* Fix tests

* Test prefetch in setup

* Add start of cache

* try add diff cache options

* Add hacky testcase

* not needed

* allow the use of a percentage for prefetch

If the TTL falls below xx% do a prefetch, if the record was popular.
Some other fixes and correctly prefetch only popular records.

1 files changed, 129 insertions, 0 deletions

diff --git a/middleware/pkg/cache/cache.go b/middleware/pkg/cache/cache.go
new file mode 100644
index 000000000..56cae2180
--- /dev/null
+++ b/middleware/pkg/cache/cache.go
@@ -0,0 +1,129 @@
+// Package cache implements a cache. The cache hold 256 shards, each shard
+// holds a cache: a map with a mutex. There is no fancy expunge algorithm, it
+// just randomly evicts elements when it gets full.
+package cache
+
+import (
+	"hash/fnv"
+	"sync"
+)
+
+// Hash returns the FNV hash of what.
+func Hash(what []byte) uint32 {
+	h := fnv.New32()
+	h.Write(what)
+	return h.Sum32()
+}
+
+// Cache is cache.
+type Cache struct {
+	shards [shardSize]*shard
+}
+
+// shard is a cache with random eviction.
+type shard struct {
+	items map[uint32]interface{}
+	size  int
+
+	sync.RWMutex
+}
+
+// New returns a new cache.
+func New(size int) *Cache {
+	ssize := size / shardSize
+	if ssize < 512 {
+		ssize = 512
+	}
+
+	c := &Cache{}
+
+	// Initialize all the shards
+	for i := 0; i < shardSize; i++ {
+		c.shards[i] = newShard(ssize)
+	}
+	return c
+}
+
+// Add adds a new element to the cache. If the element already exists it is overwritten.
+func (c *Cache) Add(key uint32, el interface{}) {
+	shard := key & (shardSize - 1)
+	c.shards[shard].Add(key, el)
+}
+
+// Get looks up element index under key.
+func (c *Cache) Get(key uint32) (interface{}, bool) {
+	shard := key & (shardSize - 1)
+	return c.shards[shard].Get(key)
+}
+
+// Remove removes the element indexed with key.
+func (c *Cache) Remove(key uint32) {
+	shard := key & (shardSize - 1)
+	c.shards[shard].Remove(key)
+}
+
+// Len returns the number of elements in the cache.
+func (c *Cache) Len() int {
+	l := 0
+	for _, s := range c.shards {
+		l += s.Len()
+	}
+	return l
+}
+
+// newShard returns a new shard with size.
+func newShard(size int) *shard { return &shard{items: make(map[uint32]interface{}), size: size} }
+
+// Add adds element indexed by key into the cache. Any existing element is overwritten
+func (s *shard) Add(key uint32, el interface{}) {
+	l := s.Len()
+	if l+1 > s.size {
+		s.Evict()
+	}
+
+	s.Lock()
+	s.items[key] = el
+	s.Unlock()
+}
+
+// Remove removes the element indexed by key from the cache.
+func (s *shard) Remove(key uint32) {
+	s.Lock()
+	delete(s.items, key)
+	s.Unlock()
+}
+
+// Evict removes a random element from the cache.
+func (s *shard) Evict() {
+	s.Lock()
+	defer s.Unlock()
+
+	key := -1
+	for k := range s.items {
+		key = int(k)
+		break
+	}
+	if key == -1 {
+		// empty cache
+		return
+	}
+	delete(s.items, uint32(key))
+}
+
+// Get looks up the element indexed under key.
+func (s *shard) Get(key uint32) (interface{}, bool) {
+	s.RLock()
+	el, found := s.items[key]
+	s.RUnlock()
+	return el, found
+}
+
+// Len returns the current length of the cache.
+func (s *shard) Len() int {
+	s.RLock()
+	l := len(s.items)
+	s.RUnlock()
+	return l
+}
+
+const shardSize = 256