Allocator Wrestling

Transcript

1. Allocator Wrestling
   Eben Freeman
   @_emfree_
   

   View Slide

2. currently: building things at honeycomb.io
   Hi everyone!
   I'm Eben
   these slides:
   speakerdeck.com/emfree/allocator-wrestling
   

   View Slide

3. Why this talk
   - Go is a managed-memory language
   - The runtime does a lot of sophisticated work on behalf of you, the programmer
   

   View Slide

4. Why this talk
   - Go is a managed-memory language
   - The runtime does a lot of sophisticated work on behalf of you, the programmer
   - And yet, dynamic memory allocation is not free
   - A program's allocation patterns can substantially affect its performance!
   

   View Slide

5. A motivating example
   Storage / query service at the day job
   (Honeycomb):
   - goal: <5 second query latency
   - up to billions of rows per query
   - flexible queries,
   on-the-fly schema changes
   

   View Slide

6. A motivating example
   A few rounds of just memory-efficiency optimization: 2-3x speedup
   

   View Slide

7. Why this talk
   - A program's allocation patterns can substantially affect its performance!
   - However, those patterns are often syntactically opaque.
   - Equipped with understanding and the right set of tools,
   we can spend our time more wisely.
   

   View Slide

8. Why this talk
   - A program's allocation patterns can substantially affect its performance!
   - However, those patterns are often syntactically opaque.
   - Equipped with understanding and the right set of tools,
   we can spend our time more wisely.
   - Moreover, the runtime's internals are inherently interesting!
   

   View Slide

9. Outline
   I. To tame the beast, you must first understand its mind
   How do the allocator and garbage collector work?
   II. Bring your binoculars
   Tools for understanding allocation patterns
   III. Delicious treats for the ravenous allocator
   Strategies for improving memory efficiency
   

   View Slide

10. This is a practitioner's perspective.
    The Go runtime authors are much smarter
    than me, maybe even smarter than you!
    And they're always cooking up new stuff
    A Caveat
    The Go runtime team discusses the design of the garbage collector.
    

    View Slide

11. I. Allocator Internals
    

    View Slide

12. Memory Layout
    func f() *int {
    // ...
    x := 22
    y := 44
    // ...
    return &y
    }
    Depending on their lifetimes, objects are allocated either on stacks or on the heap.
    

    View Slide

13. Memory Layout
    func f() *int {
    // ...
    x := 22
    y := 44
    // ...
    return &y
    }
    Depending on their lifetimes, objects are allocated either on stacks or on the heap.
    

    View Slide

14. Memory Layout
    Design goals for an allocator:
    - Efficiently satisfy allocations of a given size, but avoid fragmentation
    - Avoid locking in the common case
    - Efficiently reclaim freeable memory
    

    View Slide

15. Memory Layout
    Design goals for an allocator:
    - Efficiently satisfy allocations of a given size, but avoid fragmentation
    allocate like-sized objects in blocks
    - Avoid locking in the common case
    - Efficiently reclaim freeable memory
    

    View Slide

16. Memory Layout
    Design goals for an allocator:
    - Efficiently satisfy allocations of a given size, but avoid fragmentation
    allocate like-sized objects in blocks
    - Avoid locking in the common case
    maintain local caches
    - Efficiently reclaim freeable memory
    

    View Slide

17. Memory Layout
    Design goals for an allocator:
    - Efficiently satisfy allocations of a given size, but avoid fragmentation
    allocate like-sized objects in blocks
    - Avoid locking in the common case
    maintain local caches
    - Efficiently reclaim freeable memory
    use bitmaps for metadata, run GC concurrently
    

    View Slide

18. Memory Layout
    The heap is divided into two levels of structure: arenas and spans.
    

    View Slide

19. Memory Layout
    The heap is divided into two levels of structure: arenas and spans.
    Arenas are big chunks of aligned memory.
    On amd64, each arena is 64MB, so 4 million arenas cover the address space,
    and we keep track of them in a big global array (mheap.arenas)
    

    View Slide

20. So if a stranger on the street hands us a (valid) pointer,
    we can easily find its heap metadata!
    heapArena := mheap_.arenas[ptr / arenaSize]
    span := heapArena.spans[(ptr % arenaSize) / pageSize]
    stacks of
    pointers
    

    View Slide

21. What's a span?
    Managing the heap at arena granularity isn't practical, so heap objects live in spans.
    Small objects (<=32KB) live in spans of a fixed size class.
    type span struct {
    startAddr uintptr
    npages uintptr
    spanclass spanClass
    // allocated/free bitmap
    allocBits *gcBits
    // ...
    }
    

    View Slide

22. What's a span?
    Managing the heap at arena granularity isn't practical, so heap objects live in spans.
    Small objects (<=32KB) live in spans of a fixed size class.
    - There are ~70 size classes
    - Their spans are 8KB-64KB
    - So we can compactly allocate small objects
    with at most a few MB overhead
    

    View Slide

23. Memory Layout
    Each P has an mcache holding a span of each size class.
    Ideally, allocations can be satisfied directly out of the mcache.
    

    View Slide

24. Memory Layout
    To allocate, we find the first free object in our cached mspan, then return its address.
    type mspan struct {
    startAddr uintpr
    freeIndex uintptr // first possibly free slot
    allocCache uint64 // used/free bitmap cache:
    // ...
    }
    

    View Slide

25. Memory Layout
    To allocate, we find the first free object in our cached mspan, then return its address.
    type mspan struct {
    startAddr uintpr
    freeIndex uintptr // first possibly free slot
    allocCache uint64 // used/free bitmap cache:
    // ...
    }
    

    View Slide

26. Memory Layout
    This means that "most" memory allocations are fast:
    1. Find a cached span with the right size (mcache.mspan[sizeClass])
    (if there's none, get a new span, cache it)
    2. Find the next free object in the span
    3. If necessary, update the heap bitmap
    (so the garbage collector knows which fields are pointers)
    and require no locking!
    

    View Slide

27. Memory Layout
    What have we got so far?
    ✓ Efficiently satisfy allocations of a given size, but avoid fragmentation
    allocate like-sized objects in blocks
    ✓ Avoid locking in the common case
    maintain local caches
    ? Efficiently reclaim free memory
    

    View Slide

28. Memory Layout
    What have we got so far?
    ✓ Efficiently satisfy allocations of a given size, but avoid fragmentation
    allocate like-sized objects in blocks
    ✓ Avoid locking in the common case
    maintain local caches
    ? Efficiently reclaim free memory
    What about garbage collection?
    

    View Slide

29. Garbage collection
    We have to find and reclaim objects
    once they're no longer referenced.
    checksum(filename) uint64 {
    data := read(filename)
    // compute checksum
    }
    func read(filename) []byte {
    // open and read file
    }
    checksum
    read
    

    View Slide

30. Garbage collection
    We have to find and reclaim objects
    once they're no longer referenced.
    checksum(filename) uint64 {
    data := read(filename)
    // compute checksum
    }
    func read(filename) []byte {
    // open and read file
    }
    checksum
    

    View Slide

31. Garbage collection
    We have to find and reclaim objects
    once they're no longer referenced.
    

    View Slide

32. Garbage collection
    We have to find and reclaim objects
    once they're no longer referenced.
    Go uses a tricolor concurrent mark-sweep
    garbage collector.
    

    View Slide

33. Garbage collection
    We have to find and reclaim objects
    once they're no longer referenced.
    Go uses a tricolor concurrent mark-sweep
    garbage collector.
    GC is divided into (roughly) two phases:
    MARK: find reachable (live) objects
    (this is where the action happens)
    SWEEP: free unreachable objects
    

    View Slide

34. Garbage collection
    In the mark phase,
    objects are white, grey, or black.
    Initially, all objects are white.
    We start by marking goroutine stacks
    and globals.
    

    View Slide

35. Garbage collection
    In the mark phase,
    objects are white, grey, or black.
    Initially, all objects are white.
    We start by marking goroutine stacks
    and globals.
    When we reach an object, we mark it grey.
    

    View Slide

36. Garbage collection
    In the mark phase,
    objects are white, grey, or black.
    Initially, all objects are white.
    We start by marking goroutine stacks
    and globals.
    When we reach an object, we mark it grey.
    When an object's referents are all marked,
    we mark it black.
    

    View Slide

37. Garbage collection
    In the mark phase,
    objects are white, grey, or black.
    Initially, all objects are white.
    We start by marking goroutine stacks
    and globals.
    When we reach an object, we mark it grey.
    When an object's referents are all marked,
    we mark it black.
    

    View Slide

38. Garbage collection
    At the end, objects are either white or black.
    White objects can then be swept and freed.
    

    View Slide

39. Garbage collection
    Questions:
    - How do we know what an object's
    referents are?
    - How do we actually mark an object?
    

    View Slide

40. Garbage collection
    Questions:
    - How do we know what an object's
    referents are?
    - How do we actually mark an object?
    Use bitmaps for metadata!
    

    View Slide

41. Garbage collection
    Say we have something like:
    type Row struct {
    index int
    data []uint64
    }
    How does the garbage collector know what
    other objects it points to?
    I.e., which of its fields are pointers?
    

    View Slide

42. Garbage collection
    Remember that this heap object is actually inside an arena!
    The arena's bitmap tells us which of its words are pointers.
    

    View Slide

43. Garbage collection
    Similarly, mark state is kept in a span's gcMark bits
    type span struct {
    startAddr uintptr
    // ...
    // allocated/free bitmap
    allocBits *gcBits
    // mark state bitmap
    gcMarkBits *gcBits
    // ...
    }
    

    View Slide

44. Garbage collection
    Similarly, mark state is kept in a span's gcMark bits
    type span struct {
    startAddr uintptr
    // ...
    // allocated/free bitmap
    allocBits *gcBits
    // mark state bitmap
    gcMarkBits *gcBits
    // ...
    }
    
    

    View Slide

45. Garbage collection
    Similarly, mark state is kept in a span's gcMark bits
    type span struct {
    startAddr uintptr
    // ...
    // allocated/free bitmap
    allocBits *gcBits
    // mark state bitmap
    gcMarkBits *gcBits
    // ...
    }
    

    View Slide

46. Garbage collection
    Similarly, mark state is kept in a span's gcMark bits
    type span struct {
    startAddr uintptr
    // ...
    // allocated/free bitmap
    allocBits *gcBits
    // mark state bitmap
    gcMarkBits *gcBits
    // ...
    }
    

    View Slide

47. Garbage collection
    Once we're done marking, unmarked bits correspond to free slots!
    type span struct {
    startAddr uintptr
    // ...
    // allocated/free bitmap
    allocBits *gcBits
    // mark state bitmap
    gcMarkBits *gcBits
    // ...
    }
    

    View Slide

48. Garbage collection
    Once we're done marking, unmarked bits correspond to free slots!
    type span struct {
    startAddr uintptr
    // ...
    // allocated/free bitmap
    allocBits *gcBits
    // mark state bitmap
    gcMarkBits *gcBits
    // ...
    }
    

    View Slide

49. Garbage collection
    The garbage collector is concurrent . . . with a twist.
    If we're not careful, the application can do sneaky stuff to thwart the garbage
    collector.
    type S struct {
    p *int
    }
    func f(s *S) *int {
    r := s.p
    s.p = nil
    return r
    }
    s
    s.p
    

    View Slide

50. Garbage collection
    The garbage collector is concurrent . . . with a twist.
    If we're not careful, the application can do sneaky stuff to thwart the garbage
    collector.
    type S struct {
    p *int
    }
    func f(s *S) *int {
    r := s.p
    s.p = nil
    return r
    }
    s
    s.p
    r
    

    View Slide

51. Garbage collection
    The garbage collector is concurrent . . . with a twist.
    If we're not careful, the application can do sneaky stuff to thwart the garbage
    collector.
    type S struct {
    p *int
    }
    func f(s *S) *int {
    r := s.p
    s.p = nil
    return r
    }
    s
    r
    

    View Slide

52. Garbage collection
    The garbage collector is concurrent . . . with a twist.
    If we're not careful, the application can do sneaky stuff to thwart the garbage
    collector.
    type S struct {
    p *int
    }
    func f(s *S) *int {
    r := s.p
    s.p = nil
    return r
    }
    

    View Slide

53. Garbage collection
    The garbage collector is concurrent . . . with a twist.
    If we're not careful, the application can do sneaky stuff to thwart the garbage
    collector.
    Now we have a live pointer to memory that the garbage collector can free!
    

    View Slide

54. Garbage collection
    To avoid this peril, the compiler turns pointer writes into potential calls into the
    write barrier; very roughly:
    if writeBarrier.enabled {
    shade(*ptr)
    if current stack is grey {
    shade(val)
    }
    *ptr = val
    }
    *ptr = val
    

    View Slide

55. Garbage collection
    To avoid this peril, the compiler turns pointer writes into potential calls into the
    write barrier; very roughly:
    if writeBarrier.enabled {
    shade(*ptr)
    if current stack is grey {
    shade(val)
    }
    *ptr = val
    }
    *ptr = val
    

    View Slide

56. While the garbage collector is marking:
    - the write barrier is on
    - marking consumes resources
    - background marking
    - GC assist
    Garbage collection
    

    View Slide

57. Garbage collection
    During marking, 25% of GOMAXPROCS are dedicated to background marking.
    But a rapidly allocating goroutine can outrun it.
    Dedicated GC worker
    MARK ASSIST
    

    View Slide

58. Garbage collection
    So during marking, a goroutine gets charged for each allocation.
    If it's in debt, it has to do mark work before continuing.
    func mallocgc(size uintptr, ...) unsafe.Pointer {
    // ...
    assistG.gcAssistBytes -= int64(size)
    if assistG.gcAssistBytes < 0 {
    // This goroutine is in debt. Assist the GC to
    // this before allocating. This must happen
    // before disabling preemption.
    gcAssistAlloc(assistG)
    }
    // ...
    

    View Slide

59. Garbage collection
    In summary:
    - The runtime allocates data in spans to avoid fragmentation
    - Local caches speed up allocation,
    but the allocator still has to do some bookkeeping
    - GC is concurrent, but write barriers and mark assists can slow a program
    - GC work is proportional to scannable heap
    

    View Slide

60. II. Tools
    

    View Slide

61. Question
    It seems like dynamic memory allocation has some cost.
    Does this mean that reducing allocations will improve performance?
    

    View Slide

62. Question
    It seems like dynamic memory allocation has some cost.
    Does this mean that reducing allocations will improve performance?
    Well, it depends.
    

    View Slide

63. Question
    It seems like dynamic memory allocation has some cost.
    Does this mean that reducing allocations will improve performance?
    Well, it depends.
    The builtin memory profiler can tell us where we're allocating, but doesn't answer
    the causal question "will reducing allocations make a difference?"
    

    View Slide

64. Question
    It seems like dynamic memory allocation has some cost.
    Does this mean that reducing allocations will improve performance?
    Well, it depends.
    The builtin memory profiler can tell us where we're allocating, but doesn't answer
    the causal question "will reducing allocations make a difference?"
    Three tools to start with:
    - crude experimenting
    - sampling profiling with pprof
    - go tool trace
    

    View Slide

65. Crude experimenting
    We think that the allocator and garbage collector have some overhead,
    but we're not sure how much.
    Well . . .
    

    View Slide

66. Crude experimenting
    We think that the allocator and garbage collector have some overhead,
    but we're not sure how much.
    Well . . .
    

    View Slide

67. Crude experimenting
    Turn off the garbage collector or the allocator with runtime flags:
    GOGC=off : Disables garbage collector
    GODEBUG=sbrk=1 : Replaces entire allocator with simple persistent allocator
    

    View Slide

68. Crude experimenting
    ~30% speedup on some benchmarks:
    

    View Slide

69. Crude experimenting
    This might seem kind of stupid, but it's a cheap way to establish expectations.
    If we see speedup, that's a hint that we can optimize!
    

    View Slide

70. This might seem kind of stupid, but it's a cheap way to establish expectations.
    If we see speedup, that's a hint that we can optimize!
    Problems:
    - not viable in production: need synthetic benchmarks
    - persistent allocator isn't free either, so this doesn't fully reflect allocation cost
    Crude experimenting
    

    View Slide

71. Profiling
    A pprof CPU profile can often show time spent in runtime.mallocgc
    Tips:
    - Use the flamegraph viewer in the pprof web UI
    - If pprof isn't enabled in your binary, you can use Linux perf too
    

    View Slide

72. Profiling
    refilling span caches GC assist
    

    View Slide

73. Profiling
    Problems:
    - Program might not be CPU-bound
    - Allocation might not be on critical path
    

    View Slide

74. go tool trace
    The execution tracer might be the best tool at our disposal to understand the impact
    of allocating.
    The execution tracer captures very granular runtime events over a short time
    window:
    curl localhost:6060/debug/pprof/trace?seconds=5 > trace.out
    Which you can visualize in a web UI
    go tool trace trace.out
    

    View Slide

75. go tool trace
    However, it can be a bit dense.
    

    View Slide

76. go tool trace
    Remember, top-level GC doesn't mean the program is blocked,
    but what happens within GC is interesting!
    

    View Slide

77. go tool trace
    Remember, top-level GC doesn't mean the program is blocked,
    but what happens within GC is interesting!
    Dedicated GC worker
    MARK ASSIST
    

    View Slide

78. go tool trace
    If you're motivated, a CL exists that parses traces to generate
    minimum mutator utilization curves
    https://golang.org/cl/60790
    

    View Slide

79. go tool trace
    Minimum mutator utilization: over a sliding time window (1ms, 10ms, etc.),
    what was the minimum amount of resources available to mutators (goroutines
    doing work)?
    

    View Slide

80. go tool trace
    This is terrific (if you ask me) -- you can see if a production service is GC-bound
    utilization is < 75% :( utilization is ~ 100% :)
    

    View Slide

81. In Summary
    Together, benchmarks with the allocator off, CPU profiles, and execution traces give
    us a sense of:
    - whether allocation / GC are affecting performance
    - which call sites are spending a lot of time allocating
    - how throughput changes during GC.
    

    View Slide

82. III. What can we change?
    

    View Slide

83. If we've concluded that allocations are a source of inefficiency, what can we do?
    - Limit pointers
    - Allocate in batches
    - Try to recycle objects
    

    View Slide

84. What about tuning GOGC?
    - Absolutely helps with throughput! However . . .
    - If we want to optimize for throughput, GOGC doesn't express the real goal:
    "use all available memory, but no more"
    - Live heap size is generally (but not always) small
    - High GOGC makes avoiding OOMS harder
    But first
    

    View Slide

85. Limit pointers
    Sometimes, spurious heap allocations are easily avoided!
    func (c *ColumnManager) ReadRows() {
    // ...
    for !abort {
    tsRecord := tsReader.read()
    ts := time.Unix(0, tsRecord.Timestamp).UTC()
    if compareTimestamps(&ts, query.Start, query.End) {
    // ...
    }
    

    View Slide

86. Limit pointers
    Sometimes, spurious heap allocations are easily avoided!
    func (c *ColumnManager) ReadRows() {
    // ...
    for !abort {
    tsRecord := tsReader.read()
    ts := time.Unix(0, tsRecord.Timestamp).UTC()
    if compareTimestamps(&ts, query.Start, query.End) {
    // ...
    }
    Tight loop
    

    View Slide

87. Limit pointers
    Sometimes, spurious heap allocations are easily avoided!
    func (c *ColumnManager) ReadRows() {
    // ...
    for !abort {
    tsRecord := tsReader.read()
    ts := time.Unix(0, tsRecord.Timestamp).UTC()
    if compareTimestamps(&ts, query.Start, query.End) {
    // ...
    }
    Tight loop
    gratuitous pointer
    

    View Slide

88. Limit pointers
    Sometimes, spurious heap allocations are easily avoided!
    func (c *ColumnManager) ReadRows() {
    // ...
    for !abort {
    tsRecord := tsReader.read()
    ts := time.Unix(0, tsRecord.Timestamp).UTC()
    if compareTimestamps(&ts, query.Start, query.End) {
    // ...
    }
    Tight loop
    gratuitous pointer
    spurious heap allocation
    

    View Slide

89. Limit pointers
    The Go compiler can be enticed to tell you why a variable is heap-allocated:
    go build -gcflags="-m -m"
    but its output is a bit unwieldy.
    https://github.com/loov/view-annotated-file helps digest it:
    

    View Slide

90. Limit pointers
    for !abort {
    tsRecord := tsReader.read()
    ts := time.Unix(0, tsRecord.Timestamp).UTC()
    // . . .
    }
    var ts time.Time
    var tsNanos uint64
    

    View Slide

91. Limit pointers
    Not allocating structs with inner pointers helps the garbage collector too!
    func BenchmarkTimeAlloc(b *testing.B) {
    var x []time.Time
    for n := 0; n < b.N; n++ {
    x = make([]time.Time, 1024)
    }
    test.Check(b, len(x) == 1024)
    }
    func BenchmarkIntAlloc(b *testing.B) {
    var x []int64
    for n := 0; n < b.N; n++ {
    x = make([]int64, 1024)
    }
    test.Check(b, len(x) == 1024)
    }
    BenchmarkTimeAlloc-4 8880 ns/op BenchmarkIntAlloc-4 1540 ns/op
    

    View Slide

92. Limit pointers
    Why this discrepancy?
    BenchmarkTimeAlloc-4 8880 ns/op BenchmarkIntAlloc-4 1540 ns/op
    

    View Slide

93. Limit pointers
    Why this discrepancy?
    BenchmarkTimeAlloc-4 8880 ns/op BenchmarkIntAlloc-4 1540 ns/op
    type Time struct {
    wall uint64
    ext in64
    loc *Location
    }
    Sneaky time.Time conceals nefarious pointer!
    

    View Slide

94. Slab allocation
    Is it better to do one horse-sized allocation or 100 duck-sized allocations?
    

    View Slide

95. Slab allocation
    Although smaller allocs make better use of the mcache, larger allocs are faster on a
    per-byte basis.
    

    View Slide

96. Slab allocation
    Even though the fast path in the allocator is very optimized, we still need to do some
    work on every allocation:
    - prevent ourselves from being preempted
    - check if we need to assist GC
    - compute the next free slot in the mcache
    - set heap bitmap bits
    - etc.
    

    View Slide

97. Slab allocation
    In some cases, we can amortize that overhead by doing fewer, bigger allocs:
    // Allocate individual []interface{}s out of a big buffer
    type SlicePool struct {
    bigBuf []interface{}
    }
    func (s *SlicePool) GetSlice(size int) []interface{} {
    if size >= len(s.bigBuf) {
    s.bigBuf = make([]interface{}, blockSize)
    }
    res := s.bigBuf[:size]
    s.bigBuf = s.bigBuf[size:]
    return res
    }
    

    View Slide

98. Slab allocation
    In some cases, we can amortize that overhead by doing fewer, bigger allocs.
    The danger is that any live reference will keep the whole slab alive!
    buf := slicePool.GetSlice(24)
    w.Write(buf)
    if filters.Match(buf) {
    results <- buf
    }
    Generous subslice grants immortality
    to surrounding memory!
    

    View Slide

99. Slab allocation
    In some cases, we can amortize that overhead by doing fewer, bigger allocs.
    The danger is that any live reference will keep the whole slab alive!
    Also, these aren't safe for concurrent use: best for a few heavily-allocating
    goroutines
    buf := slicePool.GetSlice(24)
    w.Write(buf)
    if filters.Match(buf) {
    results <- buf
    Generous subslice grants immortality
    to surrounding memory!
    

    View Slide

100. Recycle allocations
     Optimization strategies:
     ✓ Avoid
     limit pointers, don't do dumb stuff
     ✓ Amortize
     do fewer, larger allocations
     ? Reuse
     recycle allocated memory
     

     View Slide

101. Storage engine architecture: two phases, mediated by channels
     Recycle allocations
     user query:
     COUNT, P95(duration)
     GROUP BY user_id
     WHERE duration > 0
     ORDER BY COUNT DESC
     LIMIT 10
     

     View Slide

102. Storage engine architecture: two phases, mediated by channels
     Recycle allocations
     - simple, easy to reason about
     - maximizes available parallelism
     - generates tons of garbage
     - data passed over channels
     - format not known in advance
     

     View Slide

103. Optimization: explicitly recycle allocated blocks of memory
     Recycle allocations
     

     View Slide

104. Optimization: explicitly recycle allocated blocks of memory
     Recycle allocations
     var buf RowBuffer
     select {
     case buf = <-recycled:
     default:
     buf = NewRowBuffer()
     }
     

     View Slide

105. A more sophisticated version: sync.Pool
     - maintains slices of recycled objects, sharded by CPU (with runtime support)
     - allows lock-free get/put in the fast path
     - caveat: cleared on every GC
     Danger:
     - must be very careful to zero or overwrite recycled memory
     Recycle allocations
     

     View Slide

106. In Review
     Your priorities and your mileage may vary!
     

     View Slide

107. In Review
     - The allocator and garbage collector are pretty ingenious!
     - Single allocations are fast but not free
     - The garbage collector can stop individual goroutines, even without STW
     - GC work depends on pointer density
     - Bring your toolbox:
     - benchmark with GC off
     - use CPU profiler to find hot allocations
     - use execution tracer to understand GC pattern
     - use escape analyzer to understand why allocations happen
     

     View Slide

108. Thank you!
     Special credit to Ian Wilkes and many others at
     Honeycomb: the true masterminds
     Suggested further reading:
     Allocation efficiency in high-performance Go services
     Achille Roussel / Rick Branson
     segment.com/blog/allocation-efficiency-in-high-
     performance-go-services/
     Go 1.5 concurrent garbage collector pacing
     Austin Clements
     golang.org/s/go15gcpacing
     So You Wanna Go Fast
     Tyler Treat
     bravenewgeek.com/so-you-wanna-go-fast/
     @_emfree_
     

     View Slide

109. - Optimizing for latency makes a lot of sense!
     Reliably low-latency garbage collection is essential for many use cases:
     chat, streaming, lockservers, low-latency high-fanout services, etc.
     - For throughput-oriented use cases, Go's pauseless garbage collector may not
     be theoretically optimal.
     (But it might not be an existential burden either.)
     A Caveat
     

     View Slide