Video and slides synchronized, mp3 and slide download available at URL http://bit.ly/1tfbFMp. Martin Thompson focuses on the evolution of Java and how it contrasts to C/C++, covering the cultural challenges of pushing the limits of performance and how to collaborate with industry experts and organize teams, which often stands at odds with the culture in many organisations. Filmed at qconlondon.com. Martin Thompson is a high-performance and low-latency specialist, with over two decades working with large scale transactional and big-data systems, in the automotive, gaming, financial, mobile, and CMS domains. He believes in Mechanical Sympathy, which is applying an understanding of the hardware to the creation of software, being fundamental to delivering elegant high-performance solutions.

1. Practicing at the Cutting Edge
Learning and Unlearning
about Performance
Martin Thompson - @mjpt777

2. InfoQ.com: News & Community Site
• 750,000 unique visitors/month
• Published in 4 languages (English, Chinese, Japanese and Brazilian
Portuguese)
• Post content from our QCon conferences
• News 15-20 / week
• Articles 3-4 / week
• Presentations (videos) 12-15 / week
• Interviews 2-3 / week
• Books 1 / month
Watch the video with slide
synchronization on InfoQ.com!
http://www.infoq.com/presentations
/java-evolution-c-c-plus-plus

3. Presented at QCon London
www.qconlondon.com
Purpose of QCon
- to empower software development by facilitating the spread of
knowledge and innovation
Strategy
- practitioner-driven conference designed for YOU: influencers of
change and innovation in your teams
- speakers and topics driving the evolution and innovation
- connecting and catalyzing the influencers and innovators
Highlights
- attended by more than 12,000 delegates since 2007
- held in 9 cities worldwide

4. Learning and
Unlearning

5. 1. Brief History of Java
2. Evolving Design approach
3. An evolving Hardware platform
4. Changes in Culture

6. 1.
Brief History of Java

7. 1.0
1396
Java Timeline
97 98 99 00 01 02 03 04 05 06 07 08 09 10 11 12
In the beginning Java was Oak
and Oak was slow
7651.41.31.21.1

8. How did it all start?

13. How did Java perform then?

16. ...but everything must have a
beginning

17. 1.0
1396
Java Timeline
97 98 99 00 01 02 03 04 05 06 07 08 09 10 11 12
Netscape Navigator 2.0 & IE 3.0
with Applet support
7651.41.31.21.1

18. Ma, I don’t need to worry about
deleting objects any more...

21. ...but we can distribute rich GUIs
without an installer

22. We could even do multi-threaded
apps on Windows 3.11

23. 1.0
1396
Java Timeline
97 98 99 00 01 02 03 04 05 06 07 08 09 10 11 12
Symantec and Microsoft
JIT Performance Wars
7651.41.31.21.1

25. Life with early JIT Compilers

26. ...cross platform GUIs are
a real possibility

27. 1.0
1396
Java Timeline
97 98 99 00 01 02 03 04 05 06 07 08 09 10 11 12
Generational Garbage Collection
Servlets
7651.41.31.21.1

31. But then came the surprise win

32. Servlets vs CGI

33. Graph Theory!

34. 1.0
1396
Java Timeline
97 98 99 00 01 02 03 04 05 06 07 08 09 10 11 12
Hotspot Compiler
JRockit for x86
7651.41.31.21.1

36. EJB

38. 1.0
1396
Java Timeline
97 98 99 00 01 02 03 04 05 06 07 08 09 10 11 12
Native Threads on Linux + epoll
NIO
7651.41.31.21.1

41. Java Memory Model

42. 1.0
1396
Java Timeline
97 98 99 00 01 02 03 04 05 06 07 08 09 10 11 12
Java Util Concurrent + JMM
Class Data Sharing
7651.41.31.21.1

43. Herb decided to write an essay...

44. The free lunch
is over...

45. Java finds its way into our poor
little mobile phones

48. What about life on the Server?

49. EJB

51. EJB

52. 1.0
1396
Java Timeline
97 98 99 00 01 02 03 04 05 06 07 08 09 10 11 12
Escape Analysis
Register Allocation
Split Bytecode Verification
7651.41.31.21.1

55. What about our hero who freed us
from the EJB tar pit?

60. SQL too difficult?

61. The world of hardware
undergoes big changes!

63. High Frequency Trading

64. Milliseconds Matter...

65. ...Microseconds
Matter...

66. ...right from market
open...

67. ...and especially after a
quiet period...

68. ...it is like being at War

71. Memory Cliff

72. SHM, Locks, and Multi-core

73. Safepoints!!!

74. 1.0
1396
Java Timeline
97 98 99 00 01 02 03 04 05 06 07 08 09 10 11 12
Tiered Compilation
G1 Garbage Collector
Compressed Pointers
7651.41.31.21.1

75. What does the near future hold?

77. sun.misc.Unsafe

80. What is everyone excited about?

82. What about other languages?

84. Now on with the story...

85. 2.
Evolving Design Approach

86. GUI Era

89. Battle for the Desktop

90. Battle for the Desktop

91. What was the biggest
factor in performance?

92. People

93. Comms design is
significant to GUIs

94. Users love responsive interfaces

95. Modelling

97. “Big” Data Era

98. What do the following
companies have in common?

100. Huge product catalogues

101. PCM
Sales
Manufacturing
Warehousing
Logistics Partners
e-Procurement
Websites
Catalogues

102. Parsing

103. Data Modelling & Representation

104. Indexing

108. Set Theory Rocks!

109. Databases are very powerful!

110. Immutable Data Rocks!

111. JDBC Drivers Suck!!!

112. Stream Processing Rocks!

113. public void characters(char[] ch,
int start,
int length)
throws SAXException

114. public void characters(char[] ch,
int start,
int length)
throws SAXException
public void startElement(String uri,
String localName,
String qName,
Attributes atts)
throws SAXException

115. public String[] split(String regex)

116. public String[] split(String regex)
public Iterable<String> split(String regex)

117. public String[] split(String regex)
public Iterable<String> split(String regex)
public void split(String regex,
Collection<String> dst)

118. Modelling

119. OOP Set Theory
Design Paradigms

120. Transaction
Processing
Era

122. Synchronous designs
are seriously limited

123. Staged Event Driven Architecture

124. The importance of Latency

125. Latency given Throughput
Latency
Load

126. The issues with Concurrency

127. The Disruptor

128. The value of Determinism

129. The value of pure Functions

130. Modelling

131. OOP Set Theory
Design Paradigms
FP

132. 3.
Evolving Hardware
Platform

133. Mechanical Sympathy

134. All Storage is Tape

135. Networks are faster than
Storage

136. Memory Access Patterns Matter

137. What if you could build a HashMap
with reified generics
and arrays of structures?

138. Buckets

139. A1
Bob
Hash
Buckets
Key Value Next

140. A1
Bob
Hash
Buckets
FF Jim
Key Value Next
HashKey Value Next

141. A1
Bob
Hash
Buckets
FF Jim
Key Value Next
HashKey Value Next
C7
Fred
HashKey Value Next

142. Buckets Key Value Hash Next

143. 0
A1 Bob 4 -1
Buckets Key Value Hash Next

144. 1
0
A1 Bob 4 -1
Buckets Key Value Hash Next
C7 Fred 2 -1

145. 1
0
A1 Bob 4 2
Buckets Key Value Hash Next
C7 Fred 2 -1
FF Jim 4 -1

146. .Net Dictionary is 10X faster
than HashMap for 2GB of data

147. ObjectLayout
PackedObjects
Value Types (Arrays 2.064)

148. “Functional Programming will solve
the multi-core problem”
- Functional Fanboi

149. Persistent Data Structures

150. Market
HAMT Ref
*
HAMT
Venue
*
HAMT
Instrument
*
HAMT
Order
*

151. Market
HAMT Ref
*
HAMT
Venue
*
HAMT
Instrument
*
HAMT
Order
*
<< CAS Failure? >>

152. Market
HAMT Ref
*
HAMT
Venue
*
HAMT
Instrument
*
HAMT
Order
* ModelDepth
Threads
<< CAS Failure? >>

153. Shared Nothing to Scale
Think Transputers
but with message
passing via SHM

154. 4.
Changes in Culture

156. Measure Everything

159. Questions?
Blog: http://mechanical-sympathy.blogspot.com/
Twitter: @mjpt777
“It does not matter how intelligent you are,
if you guess and that guess cannot be backed
up by experimental evidence -
then it is still a guess.”
- Richard Feynman

160. Watch the video with slide synchronization on
InfoQ.com!
http://www.infoq.com/presentations/java-
evolution-c-c-plus-plus