Tradeoffs, bad science, and polar bears: the world of Java optimisation

Welcome to the Java optimisation jungle. Why can’t we “just make it go faster”? It turns out, in most cases, we need to first work out “faster for whom?” and “why do we want to go faster?” and “what even *is* faster?” This talk introduces the basic principles of optimisation, before bouncing through the pitfalls of optimisation; why the exact same techniques which make Quarkus rocket-fast used to be a terrible idea fifteen years ago, why fast benchmarks make for slow programs, and why even though it can be easy to get wrong, optimisation really really matters. Along the way we’ll talk about measuring things, bad advice, garbage collection, and climate change.

Transcript

1. tradeoffs, bad science, and polar bears:
   
   
   the world of java optimisation
   
   
   Holly Cummins
   
   
   Red Hat
   
   
   

   View Slide

2. #Quarkus @holly_cummins
   why optimise?
   

   View Slide

3. #Quarkus @holly_cummins
   why optimise?
   

   View Slide

4. #Quarkus @holly_cummins
   why optimise?
   0.5s extra search
   page time
   

   View Slide

5. #Quarkus @holly_cummins
   why optimise?
   0.5s extra search
   page time
   20% drop in
   traffic
   

   View Slide

6. #Quarkus @holly_cummins
   why optimise?
   0.5s extra search
   page time
   20% drop in
   traffic
   100 ms latency on page
   load
   

   View Slide

7. #Quarkus @holly_cummins
   why optimise?
   0.5s extra search
   page time
   20% drop in
   traffic
   100 ms latency on page
   load
   7% lower
   conversion rate
   

   View Slide

8. #Quarkus @holly_cummins
   why optimise?
   0.5s extra search
   page time
   20% drop in
   traffic
   100 ms latency on page
   load
   7% lower
   conversion rate
   

   View Slide

9. #Quarkus @holly_cummins
   why optimise?
   0.5s extra search
   page time
   20% drop in
   traffic
   10 ms delay in trading
   platform
   100 ms latency on page
   load
   7% lower
   conversion rate
   

   View Slide

10. #Quarkus @holly_cummins
    why optimise?
    0.5s extra search
    page time
    20% drop in
    traffic
    10 ms delay in trading
    platform
    10% drop in
    revenue
    100 ms latency on page
    load
    7% lower
    conversion rate
    

    View Slide

11. #Quarkus @holly_cummins
    what is optimising?
    

    View Slide

12. #Quarkus @holly_cummins
    for whom?
    
    
    when?
    
    
    doing what?
    “make it go faster”
    

    View Slide

13. #Quarkus @holly_cummins
    human-centric (technical) design
    

    View Slide

14. #Quarkus @holly_cummins
    

    View Slide

15. #Quarkus @holly_cummins
    performance can be:
    

    View Slide

16. #Quarkus @holly_cummins
    performance can be:
    throughput
    

    View Slide

17. #Quarkus @holly_cummins
    performance can be:
    throughput
    transactions
    per second
    

    View Slide

18. #Quarkus @holly_cummins
    performance can be:
    throughput
    latency
    transactions
    per second
    

    View Slide

19. #Quarkus @holly_cummins
    performance can be:
    throughput
    latency
    transactions
    per second
    start-up
    time
    

    View Slide

20. #Quarkus @holly_cummins
    performance can be:
    throughput
    latency
    transactions
    per second
    response
    time
    start-up
    time
    

    View Slide

21. #Quarkus @holly_cummins
    performance can be:
    throughput
    latency
    ramp-up
    time
    transactions
    per second
    response
    time
    start-up
    time
    

    View Slide

22. #Quarkus @holly_cummins
    performance can be:
    throughput
    latency
    capacity
    ramp-up
    time
    transactions
    per second
    response
    time
    start-up
    time
    

    View Slide

23. #Quarkus @holly_cummins
    performance can be:
    throughput
    latency
    capacity
    ramp-up
    time
    transactions
    per second
    response
    time
    start-up
    time
    bandwidth
    

    View Slide

24. #Quarkus @holly_cummins
    performance can be:
    throughput
    latency
    capacity
    ramp-up
    time
    transactions
    per second
    response
    time
    start-up
    time
    footprint
    bandwidth
    

    View Slide

25. #Quarkus @holly_cummins
    performance can be:
    throughput
    latency
    capacity
    ramp-up
    time
    transactions
    per second
    response
    time
    start-up
    time
    CPU usage
    footprint
    bandwidth
    

    View Slide

26. #Quarkus @holly_cummins
    performance can be:
    throughput
    latency
    capacity
    utilisation
    ramp-up
    time
    transactions
    per second
    response
    time
    start-up
    time
    CPU usage
    footprint
    bandwidth
    

    View Slide

27. #Quarkus @holly_cummins
    performance can be:
    throughput
    latency
    capacity
    utilisation
    …
    ramp-up
    time
    transactions
    per second
    response
    time
    start-up
    time
    CPU usage
    footprint
    bandwidth
    

    View Slide

28. #Quarkus @holly_cummins
    performance can be:
    throughput
    latency
    capacity
    utilisation
    …
    ramp-up
    time
    transactions
    per second
    response
    time
    start-up
    time
    CPU usage
    footprint
    bandwidth
    

    View Slide

29. #Quarkus @holly_cummins
    Never underestimate the bandwidth [throughput] of a
    station wagon full of tapes hurtling down the highway.
    
    
    –Andrew Tanenbaum, 1981
    

    View Slide

30. #Quarkus @holly_cummins
    Never underestimate the bandwidth [throughput] of a
    station wagon full of tapes hurtling down the highway.
    
    
    –Andrew Tanenbaum, 1981
    but the latency is terrible …
    
    
    

    View Slide

31. @holly_cummins
    requirements change
    

    View Slide

32. @holly_cummins
    

    View Slide

33. @holly_cummins
    

    View Slide

34. @holly_cummins
    

    View Slide

35. @holly_cummins
    

    View Slide

36. @holly_cummins
    I am not
    designed for this.
    

    View Slide

37. @holly_cummins
    the world changes
    

    View Slide

38. #Quarkus @holly_cummins
    

    View Slide

39. #Quarkus @holly_cummins
    -Xmx == $
    

    View Slide

40. #Quarkus @holly_cummins
    -Xmx == $
    footprint
    

    View Slide

41. #Quarkus @holly_cummins
    

    View Slide

42. #Quarkus @holly_cummins
    which performs better?
    

    View Slide

43. #Quarkus @holly_cummins
    quarkus
    trading-off flexibility against
    startup speed and footprint
    

    View Slide

44. #Quarkus @holly_cummins
    quarkus
    trading-off flexibility against
    startup speed and footprint
    uhh … are you supposed
    to shut down applications
    after using them?
    

    View Slide

45. #Quarkus @holly_cummins
    which is better?
    

    View Slide

46. #Quarkus @holly_cummins
    which is better?
    which is faster?
    

    View Slide

47. #Quarkus @holly_cummins
    which is better?
    GraalVM
    Quarkus
    Application
    which is faster?
    

    View Slide

48. #Quarkus @holly_cummins
    which is better?
    OpenJDK GraalVM
    Quarkus Quarkus
    Application Application
    which is faster?
    

    View Slide

49. #Quarkus @holly_cummins
    which is better?
    ephemeral or serverless
    OpenJDK GraalVM
    Quarkus Quarkus
    Application Application
    which is faster?
    

    View Slide

50. #Quarkus @holly_cummins
    which is better?
    ephemeral or serverless
    OpenJDK GraalVM
    Quarkus Quarkus
    Application Application
    running your application
    for a long time
    which is faster?
    

    View Slide

51. #Quarkus @holly_cummins
    30% of VMs are zombies (antithesisgroup.com)
    behaviour at idle
    

    View Slide

52. #Quarkus @holly_cummins
    how to optimise?
    

    View Slide

53. #Quarkus @holly_cummins
    find the bottleneck.
    fix it.
    

    View Slide

54. #IBM @holly_cummins
    what could possibly go wrong?!
    

    View Slide

55. #RedHat @holly_cummins
    intuition
    pitfall 1
    

    View Slide

56. #Quarkus @holly_cummins
    this is not the place for ideas
    

    View Slide

57. #Quarkus @holly_cummins
    measure, don’t guess.
    

    View Slide

58. #Quarkus @holly_cummins
    measure the right thing
    

    View Slide

59. #Quarkus @holly_cummins
    measure the right thing
    what do your users care about?
    

    View Slide

60. #RedHat @holly_cummins
    numbers
    pitfall 2
    (The McNamara Fallacy)
    

    View Slide

61. #Quarkus @holly_cummins
    

    View Slide

62. #Quarkus @holly_cummins
    leading indicators
    

    View Slide

63. #Quarkus @holly_cummins
    leading indicators lagging indicators
    

    View Slide

64. #Quarkus @holly_cummins
    leading indicators
    we care about them
    lagging indicators
    

    View Slide

65. #Quarkus @holly_cummins
    leading indicators
    we care about them
    easy to measure
    lagging indicators
    

    View Slide

66. #Quarkus @holly_cummins
    leading indicators
    we care about them
    easy to measure
    hard to change
    lagging indicators
    

    View Slide

67. #Quarkus @holly_cummins
    leading indicators
    we care about them
    easy to measure
    hard to change
    lagging indicators
    easy to change
    

    View Slide

68. #Quarkus @holly_cummins
    leading indicators
    we care about them
    easy to measure
    hard to change
    lagging indicators
    predictive of a thing we care about
    easy to change
    

    View Slide

69. #Quarkus @holly_cummins
    leading indicators
    we care about them
    easy to measure
    hard to change
    lagging indicators
    predictive of a thing we care about
    hard to identify
    easy to change
    

    View Slide

70. #Quarkus @holly_cummins
    leading indicators
    we care about them
    easy to measure
    hard to change
    lagging indicators
    predictive of a thing we care about
    hard to identify
    easy to change
    

    View Slide

71. #Quarkus @holly_cummins
    caution:
    
    
    performance experiments for entertainment purposes only.
    
    
    do not try these at home.
    

    View Slide

72. #Quarkus @holly_cummins
    2007
    

    View Slide

73. #Quarkus @holly_cummins
    bad-ish advice: “reduce time spent in garbage collection”
    

    View Slide

74. #Quarkus @holly_cummins
    bad-ish advice: “reduce time spent in garbage collection”
    actually, garbage collection can make your application go faster
    

    View Slide

75. #Quarkus @holly_cummins
    2007
    

    View Slide

76. #Quarkus @holly_cummins
    2007
    

    View Slide

77. #Quarkus @holly_cummins
    2021
    

    View Slide

78. #Quarkus @holly_cummins
    2021
    

    View Slide

79. #Quarkus @holly_cummins
    -verbose:gc
    -Xverbosegclog:gclog.xml
    -Xcompactgc
    

    View Slide

80. #Quarkus @holly_cummins
    -verbose:gc
    -Xverbosegclog:gclog.xml
    -Xgcpolicy:optthruput
    -Xcompactgc
    

    View Slide

81. #Quarkus @holly_cummins
    -verbose:gc
    -Xverbosegclog:gclog.xml
    -Xgcpolicy:optthruput
    -Xmx110m
    -Xms110m
    -Xnocompactgc
    

    View Slide

82. #Quarkus @holly_cummins
    -verbose:gc
    -Xverbosegclog:gclog.xml
    -Xgcpolicy:optthruput
    -Xmx160m
    -Xms160m
    -Xnocompactgc
    

    View Slide

83. #Quarkus @holly_cummins
    -verbose:gc
    -Xverbosegclog:gclog.xml
    -Xgcpolicy:optthruput
    -Xmx300m
    -Xms300m
    -Xcompactgc
    why does the
    performance stay exactly the
    same no matter what gc
    settings I choose?
    

    View Slide

84. #Quarkus @holly_cummins
    by the way, this is cheating.
    (remember the ‘bad science’?)
    

    View Slide

85. #Quarkus @holly_cummins
    -verbose:gc
    -Xverbosegclog:gclog.xml
    -Xgcpolicy:optthruput
    

    View Slide

86. #Quarkus @holly_cummins
    -verbose:gc
    -Xverbosegclog:gclog.xml
    -Xgcpolicy:optthruput
    tool: GCMV
    

    View Slide

87. #Quarkus @holly_cummins
    -verbose:gc
    -Xverbosegclog:gclog.xml
    -Xgcpolicy:optthruput
    total GC time: 21.6s
    tool: GCMV
    

    View Slide

88. #Quarkus @holly_cummins
    -verbose:gc
    -Xverbosegclog:gclog.xml
    -Xgcpolicy:optthruput
    total GC time: 21.6s
    4.1% of time in GC pause
    tool: GCMV
    

    View Slide

89. #Quarkus @holly_cummins
    -verbose:gc
    -Xverbosegclog:gclog.xml
    -Xgcpolicy:optthruput
    total GC time: 21.6s
    4.1% of time in GC pause
    total GC time: 12.0s
    tool: GCMV
    

    View Slide

90. #Quarkus @holly_cummins
    -verbose:gc
    -Xverbosegclog:gclog.xml
    -Xgcpolicy:optthruput
    total GC time: 21.6s
    4.1% of time in GC pause
    total GC time: 12.0s
    3.6% of time in GC pause
    tool: GCMV
    

    View Slide

91. #Quarkus @holly_cummins
    -verbose:gc
    -Xverbosegclog:gclog.xml
    -Xgcpolicy:optthruput
    total GC time: 21.6s
    4.1% of time in GC pause
    23.9 GB garbage collected
    total GC time: 12.0s
    3.6% of time in GC pause
    13.0 GB garbage collected
    tool: GCMV
    

    View Slide

92. #Quarkus @holly_cummins
    -verbose:gc
    -Xverbosegclog:gclog.xml
    -Xgcpolicy:optthruput
    total GC time: 21.6s
    4.1% of time in GC pause
    23.9 GB garbage collected
    493 transactions/s
    total GC time: 12.0s
    3.6% of time in GC pause
    13.0 GB garbage collected
    260 transactions/s
    tool: GCMV
    

    View Slide

93. #Quarkus @holly_cummins
    -verbose:gc
    -Xverbosegclog:gclog.xml
    -Xgcpolicy:optthruput
    total GC time: 21.6s
    4.1% of time in GC pause
    23.9 GB garbage collected
    493 transactions/s
    total GC time: 12.0s
    3.6% of time in GC pause
    13.0 GB garbage collected
    260 transactions/s
    tool: GCMV
    

    View Slide

94. #Quarkus @holly_cummins
    -verbose:gc
    -Xverbosegclog:gclog.xml
    -Xgcpolicy:optthruput
    total GC time: 21.6s
    4.1% of time in GC pause
    23.9 GB garbage collected
    493 transactions/s
    total GC time: 12.0s
    3.6% of time in GC pause
    13.0 GB garbage collected
    260 transactions/s
    tool: GCMV
    

    View Slide

95. #Quarkus @holly_cummins
    total GC time: 21.6s
    4.1% of time in GC pause
    23.9 GB garbage collected
    493 transactions/s
    total GC time: 12.0s
    3.6% of time in GC pause
    13.0 GB garbage collected
    260 transactions/s
    

    View Slide

96. #Quarkus @holly_cummins
    total GC time: 21.6s
    4.1% of time in GC pause
    23.9 GB garbage collected
    493 transactions/s
    total GC time: 12.0s
    3.6% of time in GC pause
    13.0 GB garbage collected
    260 transactions/s
    

    View Slide

97. #Quarkus @holly_cummins
    total GC time: 21.6s
    4.1% of time in GC pause
    23.9 GB garbage collected
    493 transactions/s
    total GC time: 12.0s
    3.6% of time in GC pause
    13.0 GB garbage collected
    260 transactions/s
    leading indicator
    

    View Slide

98. #Quarkus @holly_cummins
    total GC time: 21.6s
    4.1% of time in GC pause
    23.9 GB garbage collected
    493 transactions/s
    total GC time: 12.0s
    3.6% of time in GC pause
    13.0 GB garbage collected
    260 transactions/s
    leading indicator
    

    View Slide

99. #Quarkus @holly_cummins
    total GC time: 21.6s
    4.1% of time in GC pause
    23.9 GB garbage collected
    493 transactions/s
    total GC time: 12.0s
    3.6% of time in GC pause
    13.0 GB garbage collected
    260 transactions/s
    leading indicator
    lagging indicator
    

    View Slide

100. #Quarkus @holly_cummins
     total GC time: 21.6s
     4.1% of time in GC pause
     23.9 GB garbage collected
     493 transactions/s
     total GC time: 12.0s
     3.6% of time in GC pause
     13.0 GB garbage collected
     260 transactions/s
     leading indicator
     lagging indicator
     ?
     

     View Slide

101. #Quarkus @holly_cummins
     total GC time: 21.6s
     4.1% of time in GC pause
     23.9 GB garbage collected
     493 transactions/s
     total GC time: 12.0s
     3.6% of time in GC pause
     13.0 GB garbage collected
     260 transactions/s
     leading indicator
     lagging indicator
     ?
     ?
     

     View Slide

102. #Quarkus @holly_cummins
     so wait, what changed to make the app faster?
     running jmeter on the same machine as the app gives a big speedup!
     

     View Slide

103. #Quarkus @holly_cummins
     “Any improvements made anywhere
     besides the bottleneck are an illusion.”
     
     
     – Gene Kim
     

     View Slide

104. #Quarkus @holly_cummins
     time kills all performance advice
     (even mine)
     

     View Slide

105. #Quarkus @holly_cummins
     the takeaways:
     
     
     gc can improve performance by rearranging the heap
     
     
     find the bottleneck
     
     
     validate advice independently
     

     View Slide

106. #RedHat @holly_cummins
     advice
     pitfall 3
     

     View Slide

107. #Quarkus @holly_cummins
     I read it on the
     internet!
     

     View Slide

108. #IBM @holly_cummins
     noooooo!
     “make one big method because method dispatching is slow”
     

     View Slide

109. #IBM @holly_cummins
     noooooo!
     “re-use your objects to help the garbage collector”
     

     View Slide

110. #IBM @holly_cummins
     noooooo!
     “to tune your JVM, use this command-line:”
     
     
     -server -Xms1g -Xmx1g -XX:PermSize=1g -XX:MaxPermSize=256m -Xmn256m
     -Xss64k -XX:SurvivorRatio=30 -XX:+UseConcMarkSweepGC -XX:
     +CMSParallelRemarkEnabled -XX:+UseCMSInitiatingOccupancyOnly
     -XX:CMSInitiatingOccupancyFraction=10 -XX:+ScavengeBeforeFullGC -XX:
     +CMSScavengeBeforeRemark -XX:+PrintGCDateStamps -verbose:gc -XX:
     +PrintGCDetails -Dsun.net.inetaddr.ttl=5 -XX:+HeapDumpOnOutOfMemoryError
     -XX:HeapDumpPath=`date`.hprof -Dcom.sun.management.jmxremote.port=5616
     -Dcom.sun.management.jmxremote.authenticate=false
     -Dcom.sun.management.jmxremote.ssl=false -server -Xms2g -Xmx2g
     -XX:MaxPermSize=256m -XX:NewRatio=1 -XX:+UseConcMarkSweepGC
     

     View Slide

111. #IBM @holly_cummins
     noooooo!
     use StringBuilder, never concatenate strings with +=
     

     View Slide

112. #IBM @holly_cummins
     noooooo!
     use StringBuilder, never concatenate strings with +=
     wait, what?
     
     
     yes, right?
     

     View Slide

113. #Quarkus @holly_cummins
     2 things ruin advice:
     
     
     • context
     
     
     • time
     

     View Slide

114. #RedHat @holly_cummins
     micro-optimisation
     pitfall 4
     

     View Slide

115. #Quarkus @holly_cummins
     

     View Slide

116. #Quarkus @holly_cummins
     static string beSlow()
     {
     string result = "";
     for (int i = 0; i < 314159; i++)
     {
     result += getStringData(i);
     }
     return result;
     }
     

     View Slide

117. #Quarkus @holly_cummins
     @Override
     public String toString() {
     String ret = "\n\tMarket Summary at: " + getSummaryDate() + “\n\t\t TSIA:" + getTSIA() + “\n\t\t openTSIA:" + getOpenTSIA()
     + “\n\t\t gain:" + getGainPercent() + “\n\t\t volume:" + getVolume();
     if ((getTopGainers() == null) || (getTopLosers() == null)) {
     return ret;
     }
     ret += "\n\t\t Current Top Gainers:";
     Iterator it = getTopGainers().iterator();
     while (it.hasNext()) {
     QuoteDataBean quoteData = it.next();
     ret += ("\n\t\t\t" + quoteData.toString());
     }
     ret += "\n\t\t Current Top Losers:";
     it = getTopLosers().iterator();
     while (it.hasNext()) {
     QuoteDataBean quoteData = it.next();
     ret += ("\n\t\t\t" + quoteData.toString());
     }
     return ret;
     }
     

     View Slide

118. #Quarkus @holly_cummins
     @Override
     public String toString() {
     String ret = "\n\tMarket Summary at: " + getSummaryDate() + “\n\t\t TSIA:" + getTSIA() + “\n\t\t openTSIA:" + getOpenTSIA()
     + “\n\t\t gain:" + getGainPercent() + “\n\t\t volume:" + getVolume();
     if ((getTopGainers() == null) || (getTopLosers() == null)) {
     return ret;
     }
     ret += "\n\t\t Current Top Gainers:";
     Iterator it = getTopGainers().iterator();
     while (it.hasNext()) {
     QuoteDataBean quoteData = it.next();
     ret += ("\n\t\t\t" + quoteData.toString());
     }
     ret += "\n\t\t Current Top Losers:";
     it = getTopLosers().iterator();
     while (it.hasNext()) {
     QuoteDataBean quoteData = it.next();
     ret += ("\n\t\t\t" + quoteData.toString());
     }
     return ret;
     }
     

     View Slide

119. #Quarkus @holly_cummins
     @Override
     public String toString() {
     String ret = "\n\tMarket Summary at: " + getSummaryDate() + “\n\t\t TSIA:" + getTSIA() + “\n\t\t openTSIA:" + getOpenTSIA()
     + “\n\t\t gain:" + getGainPercent() + “\n\t\t volume:" + getVolume();
     if ((getTopGainers() == null) || (getTopLosers() == null)) {
     return ret;
     }
     ret += "\n\t\t Current Top Gainers:";
     Iterator it = getTopGainers().iterator();
     while (it.hasNext()) {
     QuoteDataBean quoteData = it.next();
     ret += ("\n\t\t\t" + quoteData.toString());
     }
     ret += "\n\t\t Current Top Losers:";
     it = getTopLosers().iterator();
     while (it.hasNext()) {
     QuoteDataBean quoteData = it.next();
     ret += ("\n\t\t\t" + quoteData.toString());
     }
     return ret;
     }
     

     View Slide

120. #Quarkus @holly_cummins
     @Override
     public String toString() {
     String ret = "\n\tMarket Summary at: " + getSummaryDate() + “\n\t\t TSIA:" + getTSIA() + “\n\t\t openTSIA:" + getOpenTSIA()
     + “\n\t\t gain:" + getGainPercent() + “\n\t\t volume:" + getVolume();
     if ((getTopGainers() == null) || (getTopLosers() == null)) {
     return ret;
     }
     ret += "\n\t\t Current Top Gainers:";
     Iterator it = getTopGainers().iterator();
     while (it.hasNext()) {
     QuoteDataBean quoteData = it.next();
     ret += ("\n\t\t\t" + quoteData.toString());
     }
     ret += "\n\t\t Current Top Losers:";
     it = getTopLosers().iterator();
     while (it.hasNext()) {
     QuoteDataBean quoteData = it.next();
     ret += ("\n\t\t\t" + quoteData.toString());
     }
     return ret;
     }
     this never gets called
     

     View Slide

121. #Quarkus @holly_cummins
     let’s make travel energy-efficient?
     

     View Slide

122. #Quarkus @holly_cummins
     every little helps?
     

     View Slide

123. #Quarkus @holly_cummins
     every little helps?
     every optimisation is another optimisation you aren’t doing
     

     View Slide

124. #Quarkus @holly_cummins
     our platforms help
     

     View Slide

125. #Quarkus @holly_cummins
     static string beSlow()
     {
     string result = "";
     for (int i = 0; i < 314159; i++)
     {
     result += getStringData(i);
     }
     return result;
     }
     

     View Slide

126. #Quarkus @holly_cummins
     static string beSlow()
     {
     string result = “";
     result += getStringData(1);
     result += getStringData(2);
     result += getStringData(3);
     return result;
     }
     

     View Slide

127. #Quarkus @holly_cummins
     static string beSlow()
     {
     string result = “";
     result += getStringData(1);
     result += getStringData(2);
     result += getStringData(3);
     return result;
     }
     this is fine
     

     View Slide

128. #Quarkus @holly_cummins
     the JVM writers have far more time for optimising than you do
     clean, typical, code runs best
     

     View Slide

129. #Quarkus @holly_cummins
     ok, but how to optimise?
     

     View Slide

130. #Quarkus @holly_cummins
     tools
     

     View Slide

131. #Quarkus @holly_cummins
     “What you can optimize is limited to what
     you can observe.”
     
     
     -Susie Xia, Netflix
     

     View Slide

132. #Quarkus @holly_cummins
     observability
     

     View Slide

133. #Quarkus @holly_cummins
     method profiler
     
     
     GC analysis
     
     
     heap analysis
     
     
     APM
     
     
     distributed tracing
     * not free
     
     
     this is an incomplete list, because there are a lot of tools out there, and many cost money
     

     View Slide

134. #Quarkus @holly_cummins
     method profiler
     
     
     GC analysis
     
     
     heap analysis
     
     
     APM
     
     
     distributed tracing
     VisualVM
     * not free
     
     
     this is an incomplete list, because there are a lot of tools out there, and many cost money
     

     View Slide

135. #Quarkus @holly_cummins
     method profiler
     
     
     GC analysis
     
     
     heap analysis
     
     
     APM
     
     
     distributed tracing
     Mission Control
     VisualVM
     * not free
     
     
     this is an incomplete list, because there are a lot of tools out there, and many cost money
     

     View Slide

136. #Quarkus @holly_cummins
     method profiler
     
     
     GC analysis
     
     
     heap analysis
     
     
     APM
     
     
     distributed tracing
     Mission Control
     VisualVM
     * not free
     
     
     this is an incomplete list, because there are a lot of tools out there, and many cost money
     IBM Health Center
     (for OpenJ9)
     

     View Slide

137. #Quarkus @holly_cummins
     method profiler
     
     
     GC analysis
     
     
     heap analysis
     
     
     APM
     
     
     distributed tracing
     Mission Control
     flame graphs
     VisualVM
     * not free
     
     
     this is an incomplete list, because there are a lot of tools out there, and many cost money
     IBM Health Center
     (for OpenJ9)
     

     View Slide

138. #Quarkus @holly_cummins
     method profiler
     
     
     GC analysis
     
     
     heap analysis
     
     
     APM
     
     
     distributed tracing
     Mission Control
     flame graphs
     GCMV
     VisualVM
     * not free
     
     
     this is an incomplete list, because there are a lot of tools out there, and many cost money
     IBM Health Center
     (for OpenJ9)
     

     View Slide

139. #Quarkus @holly_cummins
     method profiler
     
     
     GC analysis
     
     
     heap analysis
     
     
     APM
     
     
     distributed tracing
     Mission Control
     flame graphs
     GCMV
     VisualVM
     Eclipse MAT
     * not free
     
     
     this is an incomplete list, because there are a lot of tools out there, and many cost money
     IBM Health Center
     (for OpenJ9)
     

     View Slide

140. #Quarkus @holly_cummins
     method profiler
     
     
     GC analysis
     
     
     heap analysis
     
     
     APM
     
     
     distributed tracing
     Mission Control
     flame graphs
     GCMV
     VisualVM
     Eclipse MAT
     * not free
     
     
     this is an incomplete list, because there are a lot of tools out there, and many cost money
     GlowRoot
     IBM Health Center
     (for OpenJ9)
     

     View Slide

141. #Quarkus @holly_cummins
     method profiler
     
     
     GC analysis
     
     
     heap analysis
     
     
     APM
     
     
     distributed tracing
     Mission Control
     flame graphs
     GCMV
     New Relic*
     VisualVM
     Eclipse MAT
     * not free
     
     
     this is an incomplete list, because there are a lot of tools out there, and many cost money
     GlowRoot
     IBM Health Center
     (for OpenJ9)
     

     View Slide

142. #Quarkus @holly_cummins
     method profiler
     
     
     GC analysis
     
     
     heap analysis
     
     
     APM
     
     
     distributed tracing
     Mission Control
     flame graphs
     GCMV
     New Relic*
     AppDynamics*
     VisualVM
     Eclipse MAT
     * not free
     
     
     this is an incomplete list, because there are a lot of tools out there, and many cost money
     GlowRoot
     IBM Health Center
     (for OpenJ9)
     

     View Slide

143. #Quarkus @holly_cummins
     method profiler
     
     
     GC analysis
     
     
     heap analysis
     
     
     APM
     
     
     distributed tracing
     Mission Control
     flame graphs
     GCMV
     New Relic*
     AppDynamics*
     VisualVM
     Dynatrace*
     Eclipse MAT
     * not free
     
     
     this is an incomplete list, because there are a lot of tools out there, and many cost money
     GlowRoot
     IBM Health Center
     (for OpenJ9)
     

     View Slide

144. #Quarkus @holly_cummins
     method profiler
     
     
     GC analysis
     
     
     heap analysis
     
     
     APM
     
     
     distributed tracing
     Mission Control
     flame graphs
     GCMV
     New Relic*
     AppDynamics*
     Zipkin
     VisualVM
     Dynatrace*
     Eclipse MAT
     * not free
     
     
     this is an incomplete list, because there are a lot of tools out there, and many cost money
     GlowRoot
     IBM Health Center
     (for OpenJ9)
     

     View Slide

145. #Quarkus @holly_cummins
     method profiler
     
     
     GC analysis
     
     
     heap analysis
     
     
     APM
     
     
     distributed tracing
     Mission Control
     flame graphs
     GCMV
     New Relic*
     AppDynamics*
     Jaeger
     Zipkin
     VisualVM
     Dynatrace*
     Eclipse MAT
     * not free
     
     
     this is an incomplete list, because there are a lot of tools out there, and many cost money
     GlowRoot
     IBM Health Center
     (for OpenJ9)
     

     View Slide

146. #Quarkus @holly_cummins
     Netflix microservice
     architecture
     optimising a micro-service: is that micro-optimising?
     

     View Slide

147. @holly_cummins
     you may need to know the whole system
     context to know what to optimise
     

     View Slide

148. #Quarkus @holly_cummins
     don’t forget the edges
     queueing theory helps us understand
     where the disasters happen
     

     View Slide

149. #Quarkus @holly_cummins
     “When it comes to IT performance, amateurs look
     at averages. Professionals look at distributions.”
     
     
     – Avishai Ish-Shalom
     

     View Slide

150. #Quarkus @holly_cummins
     “Nines don’t matter if your
     users aren’t happy.”
     
     
     – Charity Majors
     

     View Slide

151. #Quarkus @holly_cummins
     slow performance can turn into big cloud bills
     make cloud costs visible to engineers
     

     View Slide

152. #Quarkus @holly_cummins
     ok, but you promised bears
     

     View Slide

153. #Quarkus @holly_cummins
     if you leave the TV on when
     you’re not using it, you’re a
     polar bear murderer
     

     View Slide

154. #Quarkus @holly_cummins
     it all fits together
     

     View Slide

155. @holly_cummins #RedHat
     density
     Source: Clement Escoffier
     cost impact of framework choice
     Setup:
     
     
     • 800 requests/second, over 20 days
     
     
     • SLA > 99%
     
     
     • AWS instances
     
     
     Assumptions:
     
     
     • Costs are for us-east-1 data centre
     

     View Slide

156. @holly_cummins #RedHat
     Setup:
     
     
     • 800 requests/second, over 20 days
     
     
     • SLA > 99%
     
     
     Assumptions:
     
     
     • 50% load
     
     
     • us-east-1 data centre
     
     
     • Teads dataset
     Source: Clement Escoffier x Teads
     cloud carbon impact of framework choice
     

     View Slide

157. @holly_cummins #RedHat
     capacity
     Source: John O’Hara
     Setup:
     
     
     • REST + CRUD
     
     
     • large heap
     
     
     • RAPL energy measurement
     
     
     Assumptions:
     
     
     • US energy mix
     climate impact of framework choice
     

     View Slide

158. @holly_cummins #RedHat
     our experiments show
     strong correlations between
     

     View Slide

159. @holly_cummins #RedHat
     our experiments show
     strong correlations between
     

     View Slide

160. @holly_cummins #RedHat
     our experiments show
     strong correlations between
     • lower runtime cost
     

     View Slide

161. @holly_cummins #RedHat
     our experiments show
     strong correlations between
     • lower runtime cost
     • higher throughput
     

     View Slide

162. @holly_cummins #RedHat
     our experiments show
     strong correlations between
     • lower runtime cost
     • higher throughput
     • better carbon-efficiency
     

     View Slide

163. @holly_cummins #RedHat
     our experiments show
     strong correlations between
     • lower runtime cost
     • higher throughput
     • better carbon-efficiency
     

     View Slide

164. #Quarkus @holly_cummins
     there is a moral imperative to avoid waste
     

     View Slide

165. #Quarkus @holly_cummins
     there is a moral imperative to avoid waste
     electricity
     
     
     hardware
     

     View Slide

166. #Quarkus @holly_cummins
     data centres use 1-2% of
     the world’s electricity
     

     View Slide

167. #Quarkus @holly_cummins
     fewer devices
     
     
     longer lifetime
     

     View Slide

168. #Quarkus @holly_cummins
     fewer devices
     
     
     longer lifetime
     higher efficiency
     

     View Slide

169. #Quarkus @holly_cummins
     fewer devices
     
     
     longer lifetime
     higher efficiency
     lower footprint
     

     View Slide

170. #Quarkus @holly_cummins
     fewer devices
     
     
     longer lifetime
     higher efficiency
     lower footprint
     more multitenancy
     

     View Slide

171. #Quarkus @holly_cummins
     fewer devices
     
     
     longer lifetime
     higher efficiency
     lower footprint
     more multitenancy
     optimise for longevity
     

     View Slide

172. #Quarkus @holly_cummins
     fewer devices
     
     
     longer lifetime
     higher efficiency
     lower footprint
     more multitenancy
     optimise for longevity
     the end of planned obsolescence?
     

     View Slide

173. #Quarkus @holly_cummins
     sooo …
     
     
     you can optimise, and it can be fun
     
     
     measure, don’t guess
     
     
     only optimise what matters
     now for questions!
     
     
     @holly_cummins
     #Quarkus
     

     View Slide