swarm/feeds: Parallel feed lookups (#19414)

This commit is contained in:
Javier Peletier 2019-05-16 15:47:11 +02:00 committed by Anton Evangelatov
parent 0c5f8c078a
commit 1e067202a2
14 changed files with 936 additions and 326 deletions

View File

@ -27,7 +27,7 @@ import (
const (
hasherCount = 8
feedsHashAlgorithm = storage.SHA3Hash
defaultRetrieveTimeout = 100 * time.Millisecond
defaultRetrieveTimeout = 1000 * time.Millisecond
)
// cacheEntry caches the last known update of a specific Swarm feed.

View File

@ -23,6 +23,7 @@ import (
"context"
"fmt"
"sync"
"sync/atomic"
"github.com/ethereum/go-ethereum/swarm/chunk"
@ -178,12 +179,12 @@ func (h *Handler) Lookup(ctx context.Context, query *Query) (*cacheEntry, error)
return nil, NewError(ErrInit, "Call Handler.SetStore() before performing lookups")
}
var readCount int
var readCount int32
// Invoke the lookup engine.
// The callback will be called every time the lookup algorithm needs to guess
requestPtr, err := lookup.Lookup(ctx, timeLimit, query.Hint, func(ctx context.Context, epoch lookup.Epoch, now uint64) (interface{}, error) {
readCount++
atomic.AddInt32(&readCount, 1)
id := ID{
Feed: query.Feed,
Epoch: epoch,
@ -228,17 +229,17 @@ func (h *Handler) updateCache(request *Request) (*cacheEntry, error) {
updateAddr := request.Addr()
log.Trace("feed cache update", "topic", request.Topic.Hex(), "updateaddr", updateAddr, "epoch time", request.Epoch.Time, "epoch level", request.Epoch.Level)
feedUpdate := h.get(&request.Feed)
if feedUpdate == nil {
feedUpdate = &cacheEntry{}
h.set(&request.Feed, feedUpdate)
entry := h.get(&request.Feed)
if entry == nil {
entry = &cacheEntry{}
h.set(&request.Feed, entry)
}
// update our rsrcs entry map
feedUpdate.lastKey = updateAddr
feedUpdate.Update = request.Update
feedUpdate.Reader = bytes.NewReader(feedUpdate.data)
return feedUpdate, nil
entry.lastKey = updateAddr
entry.Update = request.Update
entry.Reader = bytes.NewReader(entry.data)
return entry, nil
}
// Update publishes a feed update

View File

@ -177,8 +177,8 @@ func TestFeedsHandler(t *testing.T) {
if err != nil {
t.Fatal(err)
}
if request.Epoch.Base() != 0 || request.Epoch.Level != 22 {
t.Fatalf("Expected epoch base time to be %d, got %d. Expected epoch level to be %d, got %d", 0, request.Epoch.Base(), 22, request.Epoch.Level)
if request.Epoch.Base() != 0 || request.Epoch.Level != 28 {
t.Fatalf("Expected epoch base time to be %d, got %d. Expected epoch level to be %d, got %d", 0, request.Epoch.Base(), 28, request.Epoch.Level)
}
data = []byte(updates[3])
request.SetData(data)
@ -213,8 +213,8 @@ func TestFeedsHandler(t *testing.T) {
if !bytes.Equal(update2.data, []byte(updates[len(updates)-1])) {
t.Fatalf("feed update data was %v, expected %v", string(update2.data), updates[len(updates)-1])
}
if update2.Level != 22 {
t.Fatalf("feed update epoch level was %d, expected 22", update2.Level)
if update2.Level != 28 {
t.Fatalf("feed update epoch level was %d, expected 28", update2.Level)
}
if update2.Base() != 0 {
t.Fatalf("feed update epoch base time was %d, expected 0", update2.Base())

View File

@ -16,11 +16,11 @@ func getTestID() *ID {
func TestIDAddr(t *testing.T) {
id := getTestID()
updateAddr := id.Addr()
compareByteSliceToExpectedHex(t, "updateAddr", updateAddr, "0x8b24583ec293e085f4c78aaee66d1bc5abfb8b4233304d14a349afa57af2a783")
compareByteSliceToExpectedHex(t, "updateAddr", updateAddr, "0x842d0a81987b9755dfeaa5558f5c134c1c0af48b6545005cac7b533d9411453a")
}
func TestIDSerializer(t *testing.T) {
testBinarySerializerRecovery(t, getTestID(), "0x776f726c64206e657773207265706f72742c20657665727920686f7572000000876a8936a7cd0b79ef0735ad0896c1afe278781ce803000000000019")
testBinarySerializerRecovery(t, getTestID(), "0x776f726c64206e657773207265706f72742c20657665727920686f7572000000876a8936a7cd0b79ef0735ad0896c1afe278781ce80300000000001f")
}
func TestIDLengthCheck(t *testing.T) {

View File

@ -0,0 +1,63 @@
package lookup
import "context"
// FluzCapacitorAlgorithm works by narrowing the epoch search area if an update is found
// going back and forth in time
// First, it will attempt to find an update where it should be now if the hint was
// really the last update. If that lookup fails, then the last update must be either the hint itself
// or the epochs right below. If however, that lookup succeeds, then the update must be
// that one or within the epochs right below.
// see the guide for a more graphical representation
func FluzCapacitorAlgorithm(ctx context.Context, now uint64, hint Epoch, read ReadFunc) (value interface{}, err error) {
var lastFound interface{}
var epoch Epoch
if hint == NoClue {
hint = worstHint
}
t := now
for {
epoch = GetNextEpoch(hint, t)
value, err = read(ctx, epoch, now)
if err != nil {
return nil, err
}
if value != nil {
lastFound = value
if epoch.Level == LowestLevel || epoch.Equals(hint) {
return value, nil
}
hint = epoch
continue
}
if epoch.Base() == hint.Base() {
if lastFound != nil {
return lastFound, nil
}
// we have reached the hint itself
if hint == worstHint {
return nil, nil
}
// check it out
value, err = read(ctx, hint, now)
if err != nil {
return nil, err
}
if value != nil {
return value, nil
}
// bad hint.
t = hint.Base()
hint = worstHint
continue
}
base := epoch.Base()
if base == 0 {
return nil, nil
}
t = base - 1
}
}

View File

@ -0,0 +1,185 @@
package lookup
import (
"context"
"sync/atomic"
"time"
)
type stepFunc func(ctx context.Context, t uint64, hint Epoch) interface{}
// LongEarthLookaheadDelay is the headstart the lookahead gives R before it launches
var LongEarthLookaheadDelay = 250 * time.Millisecond
// LongEarthLookbackDelay is the headstart the lookback gives R before it launches
var LongEarthLookbackDelay = 250 * time.Millisecond
// LongEarthAlgorithm explores possible lookup paths in parallel, pruning paths as soon
// as a more promising lookup path is found. As a result, this lookup algorithm is an order
// of magnitude faster than the FluzCapacitor algorithm, but at the expense of more exploratory reads.
// This algorithm works as follows. On each step, the next epoch is immediately looked up (R)
// and given a head start, while two parallel "steps" are launched a short time after:
// look ahead (A) is the path the algorithm would take if the R lookup returns a value, whereas
// look back (B) is the path the algorithm would take if the R lookup failed.
// as soon as R is actually finished, the A or B paths are pruned depending on the value of R.
// if A returns earlier than R, then R and B read operations can be safely canceled, saving time.
// The maximum number of active read operations is calculated as 2^(timeout/headstart).
// If headstart is infinite, this algorithm behaves as FluzCapacitor.
// timeout is the maximum execution time of the passed `read` function.
// the two head starts can be configured by changing LongEarthLookaheadDelay or LongEarthLookbackDelay
func LongEarthAlgorithm(ctx context.Context, now uint64, hint Epoch, read ReadFunc) (interface{}, error) {
if hint == NoClue {
hint = worstHint
}
var stepCounter int32 // for debugging, stepCounter allows to give an ID to each step instance
errc := make(chan struct{}) // errc will help as an error shortcut signal
var gerr error // in case of error, this variable will be set
var step stepFunc // For efficiency, the algorithm step is defined as a closure
step = func(ctxS context.Context, t uint64, last Epoch) interface{} {
stepID := atomic.AddInt32(&stepCounter, 1) // give an ID to this call instance
trace(stepID, "init: t=%d, last=%s", t, last.String())
var valueA, valueB, valueR interface{}
// initialize the three read contexts
ctxR, cancelR := context.WithCancel(ctxS) // will handle the current read operation
ctxA, cancelA := context.WithCancel(ctxS) // will handle the lookahead path
ctxB, cancelB := context.WithCancel(ctxS) // will handle the lookback path
epoch := GetNextEpoch(last, t) // calculate the epoch to look up in this step instance
// define the lookAhead function, which will follow the path as if R was successful
lookAhead := func() {
valueA = step(ctxA, t, epoch) // launch the next step, recursively.
if valueA != nil { // if this path is successful, we don't need R or B.
cancelB()
cancelR()
}
}
// define the lookBack function, which will follow the path as if R was unsuccessful
lookBack := func() {
if epoch.Base() == last.Base() {
return
}
base := epoch.Base()
if base == 0 {
return
}
valueB = step(ctxB, base-1, last)
}
go func() { //goroutine to read the current epoch (R)
defer cancelR()
var err error
valueR, err = read(ctxR, epoch, now) // read this epoch
if valueR == nil { // if unsuccessful, cancel lookahead, otherwise cancel lookback.
cancelA()
} else {
cancelB()
}
if err != nil && err != context.Canceled {
gerr = err
close(errc)
}
}()
go func() { // goroutine to give a headstart to R and then launch lookahead.
defer cancelA()
// if we are at the lowest level or the epoch to look up equals the last one,
// then we cannot lookahead (can't go lower or repeat the same lookup, this would
// cause an infinite loop)
if epoch.Level == LowestLevel || epoch.Equals(last) {
return
}
// give a head start to R, or launch immediately if R finishes early enough
select {
case <-TimeAfter(LongEarthLookaheadDelay):
lookAhead()
case <-ctxR.Done():
if valueR != nil {
lookAhead() // only look ahead if R was successful
}
case <-ctxA.Done():
}
}()
go func() { // goroutine to give a headstart to R and then launch lookback.
defer cancelB()
// give a head start to R, or launch immediately if R finishes early enough
select {
case <-TimeAfter(LongEarthLookbackDelay):
lookBack()
case <-ctxR.Done():
if valueR == nil {
lookBack() // only look back in case R failed
}
case <-ctxB.Done():
}
}()
<-ctxA.Done()
if valueA != nil {
trace(stepID, "Returning valueA=%v", valueA)
return valueA
}
<-ctxR.Done()
if valueR != nil {
trace(stepID, "Returning valueR=%v", valueR)
return valueR
}
<-ctxB.Done()
trace(stepID, "Returning valueB=%v", valueB)
return valueB
}
var value interface{}
stepCtx, cancel := context.WithCancel(ctx)
go func() { // launch the root step in its own goroutine to allow cancellation
defer cancel()
value = step(stepCtx, now, hint)
}()
// wait for the algorithm to finish, but shortcut in case
// of errors
select {
case <-stepCtx.Done():
case <-errc:
cancel()
return nil, gerr
}
if ctx.Err() != nil {
return nil, ctx.Err()
}
if value != nil || hint == worstHint {
return value, nil
}
// at this point the algorithm did not return a value,
// so we challenge the hint given.
value, err := read(ctx, hint, now)
if err != nil {
return nil, err
}
if value != nil {
return value, nil // hint is valid, return it.
}
// hint is invalid. Invoke the algorithm
// without hint.
now = hint.Base()
if hint.Level == HighestLevel {
now--
}
return LongEarthAlgorithm(ctx, now, NoClue, read)
}

View File

@ -87,5 +87,5 @@ func (e *Epoch) Equals(epoch Epoch) bool {
// String implements the Stringer interface.
func (e *Epoch) String() string {
return fmt.Sprintf("Epoch{Time:%d, Level:%d}", e.Time, e.Level)
return fmt.Sprintf("Epoch{Base: %d, Time:%d, Level:%d}", e.Base(), e.Time, e.Level)
}

View File

@ -20,7 +20,10 @@ so they can be found
*/
package lookup
import "context"
import (
"context"
"time"
)
const maxuint64 = ^uint64(0)
@ -28,8 +31,8 @@ const maxuint64 = ^uint64(0)
const LowestLevel uint8 = 0 // default is 0 (1 second)
// HighestLevel sets the lowest frequency the algorithm will operate at, as a power of 2.
// 25 -> 2^25 equals to roughly one year.
const HighestLevel = 25 // default is 25 (~1 year)
// 31 -> 2^31 equals to roughly 38 years.
const HighestLevel = 31
// DefaultLevel sets what level will be chosen to search when there is no hint
const DefaultLevel = HighestLevel
@ -43,7 +46,12 @@ type Algorithm func(ctx context.Context, now uint64, hint Epoch, read ReadFunc)
// read() will be called on each lookup attempt
// Returns an error only if read() returns an error
// Returns nil if an update was not found
var Lookup Algorithm = FluzCapacitorAlgorithm
var Lookup Algorithm = LongEarthAlgorithm
// TimeAfter must point to a function that returns a timer
// This is here so that tests can replace it with
// a mock up timer factory to simulate time deterministically
var TimeAfter = time.After
// ReadFunc is a handler called by Lookup each time it attempts to find a value
// It should return <nil> if a value is not found
@ -123,61 +131,6 @@ func GetFirstEpoch(now uint64) Epoch {
var worstHint = Epoch{Time: 0, Level: 63}
// FluzCapacitorAlgorithm works by narrowing the epoch search area if an update is found
// going back and forth in time
// First, it will attempt to find an update where it should be now if the hint was
// really the last update. If that lookup fails, then the last update must be either the hint itself
// or the epochs right below. If however, that lookup succeeds, then the update must be
// that one or within the epochs right below.
// see the guide for a more graphical representation
func FluzCapacitorAlgorithm(ctx context.Context, now uint64, hint Epoch, read ReadFunc) (value interface{}, err error) {
var lastFound interface{}
var epoch Epoch
if hint == NoClue {
hint = worstHint
}
t := now
for {
epoch = GetNextEpoch(hint, t)
value, err = read(ctx, epoch, now)
if err != nil {
return nil, err
}
if value != nil {
lastFound = value
if epoch.Level == LowestLevel || epoch.Equals(hint) {
return value, nil
}
hint = epoch
continue
}
if epoch.Base() == hint.Base() {
if lastFound != nil {
return lastFound, nil
}
// we have reached the hint itself
if hint == worstHint {
return nil, nil
}
// check it out
value, err = read(ctx, hint, now)
if err != nil {
return nil, err
}
if value != nil {
return value, nil
}
// bad hint.
t = hint.Base()
hint = worstHint
continue
}
base := epoch.Base()
if base == 0 {
return nil, nil
}
t = base - 1
}
var trace = func(id int32, formatString string, a ...interface{}) {
//fmt.Printf("Step ID #%d "+formatString+"\n", append([]interface{}{id}, a...)...)
}

File diff suppressed because one or more lines are too long

View File

@ -0,0 +1,154 @@
package lookup_test
/*
This file contains components to mock a storage for testing
lookup algorithms and measure the number of reads.
*/
import (
"context"
"fmt"
"sync"
"time"
"github.com/ethereum/go-ethereum/swarm/log"
"github.com/ethereum/go-ethereum/swarm/storage/feed/lookup"
)
// Data is a struct to keep a value to store/retrieve during testing
type Data struct {
Payload uint64
Time uint64
}
// String implements fmt.Stringer
func (d *Data) String() string {
return fmt.Sprintf("%d-%d", d.Payload, d.Time)
}
// Datamap is an internal map to hold the mocked storage
type DataMap map[lookup.EpochID]*Data
// StoreConfig allows to specify the simulated delays for each type of
// read operation
type StoreConfig struct {
CacheReadTime time.Duration // time it takes to read from the cache
FailedReadTime time.Duration // time it takes to acknowledge a read as failed
SuccessfulReadTime time.Duration // time it takes to fetch data
}
// StoreCounters will track read count metrics
type StoreCounters struct {
reads int
cacheHits int
failed int
successful int
canceled int
maxSimultaneous int
}
// Store simulates a store and keeps track of performance counters
type Store struct {
StoreConfig
StoreCounters
data DataMap
cache DataMap
lock sync.RWMutex
activeReads int
}
// NewStore returns a new mock store ready for use
func NewStore(config *StoreConfig) *Store {
store := &Store{
StoreConfig: *config,
data: make(DataMap),
}
store.Reset()
return store
}
// Reset reset performance counters and clears the cache
func (s *Store) Reset() {
s.cache = make(DataMap)
s.StoreCounters = StoreCounters{}
}
// Put stores a value in the mock store at the given epoch
func (s *Store) Put(epoch lookup.Epoch, value *Data) {
log.Debug("Write: %d-%d, value='%d'\n", epoch.Base(), epoch.Level, value.Payload)
s.data[epoch.ID()] = value
}
// Update runs the seed algorithm to place the update in the appropriate epoch
func (s *Store) Update(last lookup.Epoch, now uint64, value *Data) lookup.Epoch {
epoch := lookup.GetNextEpoch(last, now)
s.Put(epoch, value)
return epoch
}
// Get retrieves data at the specified epoch, simulating a delay
func (s *Store) Get(ctx context.Context, epoch lookup.Epoch, now uint64) (value interface{}, err error) {
epochID := epoch.ID()
var operationTime time.Duration
defer func() { // simulate a delay according to what has actually happened
select {
case <-lookup.TimeAfter(operationTime):
case <-ctx.Done():
s.lock.Lock()
s.canceled++
s.lock.Unlock()
value = nil
err = ctx.Err()
}
s.lock.Lock()
s.activeReads--
s.lock.Unlock()
}()
s.lock.Lock()
defer s.lock.Unlock()
s.reads++
s.activeReads++
if s.activeReads > s.maxSimultaneous {
s.maxSimultaneous = s.activeReads
}
// 1.- Simulate a cache read
item := s.cache[epochID]
operationTime += s.CacheReadTime
if item != nil {
s.cacheHits++
if item.Time <= now {
s.successful++
return item, nil
}
return nil, nil
}
// 2.- simulate a full read
item = s.data[epochID]
if item != nil {
operationTime += s.SuccessfulReadTime
s.successful++
s.cache[epochID] = item
if item.Time <= now {
return item, nil
}
} else {
operationTime += s.FailedReadTime
s.failed++
}
return nil, nil
}
// MakeReadFunc returns a read function suitable for the lookup algorithm, mapped
// to this mock storage
func (s *Store) MakeReadFunc() lookup.ReadFunc {
return func(ctx context.Context, epoch lookup.Epoch, now uint64) (interface{}, error) {
return s.Get(ctx, epoch, now)
}
}

View File

@ -0,0 +1,128 @@
package lookup_test
// This file contains simple time simulation tools for testing
// and measuring time-aware algorithms
import (
"sync"
"time"
)
// Timer tracks information about a simulated timer
type Timer struct {
deadline time.Time
signal chan time.Time
id int
}
// Stopwatch measures simulated execution time and manages simulated timers
type Stopwatch struct {
t time.Time
resolution time.Duration
timers map[int]*Timer
timerCounter int
stopSignal chan struct{}
lock sync.RWMutex
}
// NewStopwatch returns a simulated clock that ticks on `resolution` intervals
func NewStopwatch(resolution time.Duration) *Stopwatch {
s := &Stopwatch{
resolution: resolution,
}
s.Reset()
return s
}
// Reset clears all timers and sents the stopwatch to zero
func (s *Stopwatch) Reset() {
s.t = time.Time{}
s.timers = make(map[int]*Timer)
s.Stop()
}
// Tick advances simulated time by the stopwatch's resolution and triggers
// all due timers
func (s *Stopwatch) Tick() {
s.t = s.t.Add(s.resolution)
s.lock.Lock()
defer s.lock.Unlock()
for id, timer := range s.timers {
if s.t.After(timer.deadline) || s.t.Equal(timer.deadline) {
timer.signal <- s.t
close(timer.signal)
delete(s.timers, id)
}
}
}
// NewTimer returns a new timer that will trigger after `duration` elapses in the
// simulation
func (s *Stopwatch) NewTimer(duration time.Duration) <-chan time.Time {
s.lock.Lock()
defer s.lock.Unlock()
s.timerCounter++
timer := &Timer{
deadline: s.t.Add(duration),
signal: make(chan time.Time, 1),
id: s.timerCounter,
}
s.timers[timer.id] = timer
return timer.signal
}
// TimeAfter returns a simulated timer factory that can replace `time.After`
func (s *Stopwatch) TimeAfter() func(d time.Duration) <-chan time.Time {
return func(d time.Duration) <-chan time.Time {
return s.NewTimer(d)
}
}
// Elapsed returns the time that has passed in the simulation
func (s *Stopwatch) Elapsed() time.Duration {
return s.t.Sub(time.Time{})
}
// Run starts the time simulation
func (s *Stopwatch) Run() {
go func() {
stopSignal := make(chan struct{})
s.lock.Lock()
if s.stopSignal != nil {
close(s.stopSignal)
}
s.stopSignal = stopSignal
s.lock.Unlock()
for {
select {
case <-time.After(1 * time.Millisecond):
s.Tick()
case <-stopSignal:
return
}
}
}()
}
// Stop stops the time simulation
func (s *Stopwatch) Stop() {
s.lock.Lock()
defer s.lock.Unlock()
if s.stopSignal != nil {
close(s.stopSignal)
s.stopSignal = nil
}
}
func (s *Stopwatch) Measure(measuredFunc func()) time.Duration {
s.Reset()
s.Run()
defer s.Stop()
measuredFunc()
return s.Elapsed()
}

View File

@ -30,7 +30,7 @@ func getTestQuery() *Query {
}
func TestQueryValues(t *testing.T) {
var expected = KV{"hint.level": "25", "hint.time": "1000", "time": "5000", "topic": "0x776f726c64206e657773207265706f72742c20657665727920686f7572000000", "user": "0x876A8936A7Cd0b79Ef0735AD0896c1AFe278781c"}
var expected = KV{"hint.level": "31", "hint.time": "1000", "time": "5000", "topic": "0x776f726c64206e657773207265706f72742c20657665727920686f7572000000", "user": "0x876A8936A7Cd0b79Ef0735AD0896c1AFe278781c"}
query := getTestQuery()
testValueSerializer(t, query, expected)

View File

@ -223,7 +223,7 @@ func TestUpdateChunkSerializationErrorChecking(t *testing.T) {
t.Fatalf("error creating update chunk:%s", err)
}
compareByteSliceToExpectedHex(t, "chunk", chunk.Data(), "0x0000000000000000776f726c64206e657773207265706f72742c20657665727920686f7572000000876a8936a7cd0b79ef0735ad0896c1afe278781ce803000000000019416c206269656e206861636572206a616dc3a173206c652066616c7461207072656d696f5a0ffe0bc27f207cd5b00944c8b9cee93e08b89b5ada777f123ac535189333f174a6a4ca2f43a92c4a477a49d774813c36ce8288552c58e6205b0ac35d0507eb00")
compareByteSliceToExpectedHex(t, "chunk", chunk.Data(), "0x0000000000000000776f726c64206e657773207265706f72742c20657665727920686f7572000000876a8936a7cd0b79ef0735ad0896c1afe278781ce80300000000001f416c206269656e206861636572206a616dc3a173206c652066616c7461207072656d696f9896df5937e64e51a7994479ff3fe0ed790d539b9b3e85e93c0014a8a64374f23603c79d16e99b50a757896d3816d7022ac594ad1415679a9b164afb2e5926d801")
var recovered Request
recovered.fromChunk(chunk)

View File

@ -28,7 +28,7 @@ func getTestFeedUpdate() *Update {
}
func TestUpdateSerializer(t *testing.T) {
testBinarySerializerRecovery(t, getTestFeedUpdate(), "0x0000000000000000776f726c64206e657773207265706f72742c20657665727920686f7572000000876a8936a7cd0b79ef0735ad0896c1afe278781ce803000000000019456c20717565206c6565206d7563686f207920616e6461206d7563686f2c207665206d7563686f20792073616265206d7563686f")
testBinarySerializerRecovery(t, getTestFeedUpdate(), "0x0000000000000000776f726c64206e657773207265706f72742c20657665727920686f7572000000876a8936a7cd0b79ef0735ad0896c1afe278781ce80300000000001f456c20717565206c6565206d7563686f207920616e6461206d7563686f2c207665206d7563686f20792073616265206d7563686f")
}
func TestUpdateLengthCheck(t *testing.T) {