Price Aware Architecture

Turning Technical Debt into Monetary Debt: Price Aware Architecture

How much does your software cost to operate?

Software design and implementation has an uneasy relationship with
cost. As software developers, we often dwell on the cost of building
software, but far less so on the cost to operate. The current
zeitgeist of pay-as-you-go infrastructure makes this particularly
apparent where rigid, hard to deploy and adapt software directly
equates to larger bills. If we want to take control of operational
cost, it is critical that we start to make operational cost a
first-class constraint in the design and architecture of modern
software.

In this talk we will work through the idea of price-aware
architectures, answering: what are the properties of a price-aware
architecture? what happens when we design our systems in a way that
can leverage variable price and availability of infrastructure? and
finally, how do we implement such systems?

The talk was presented at YOW! West, May 2016.

Transcript

1. G PRICE
   AWARE
   ARCHITECTURE
   

   View Slide

2. “We are in the business of
   turning technical debt into
   monetary debt”
   - Old Ambiata Proverb
   

   View Slide

3. WHAT IS THE EFFORT TO
   BUILD?
   

   View Slide

4. WHAT IS THE COST TO
   OPERATE?
   

   View Slide

5. HOW CAN OUR SOFTWARE ADAPT TO
   COST?
   

   View Slide

6. Clouds
   

   View Slide

7. on demand
   

   View Slide

8. View Slide

9. View Slide

10. peak load
    

    View Slide

11. peak load
    

    View Slide

12. savings
    peak load
    

    View Slide

13. View Slide

14. View Slide

15. View Slide

16. excess capacity
    

    View Slide

17. savings
    peak load
    

    View Slide

18. decision time
    

    View Slide

19. variable capability
    

    View Slide

20. View Slide

21. View Slide

22. View Slide

23. View Slide

24. variable abstraction
    

    View Slide

25. View Slide

26. View Slide

27. View Slide

28. View Slide

29. excess capacity markets
    

    View Slide

30. variable price
    

    View Slide

31. View Slide

32. View Slide

33. variable availability
    

    View Slide

34. View Slide

35. View Slide

36. predictable capacity markets
    

    View Slide

37. View Slide

38. predictable
    

    View Slide

39. View Slide

40. View Slide

41. View Slide

42. transparent costs
    

    View Slide

43. View Slide

44. View Slide

45. ONE
    Know your Business
    

    View Slide

46. “find the dimension you are
    going to make money over and
    then make sure that the
    architecture follows the money”
    - Werner Vogels - 21st-Century-Architectures
    

    View Slide

47. View Slide

48. View Slide

49. View Slide

50. View Slide

51. View Slide

52. View Slide

53. $ $
    

    View Slide

54. $
    

    View Slide

55. IF IT COSTS MORE, DO YOU MAKE MORE
    MONEY?
    

    View Slide

56. budget & constraints
    

    View Slide

57. price-aware backplane
    

    View Slide

58. TWO
    Know your Bill
    

    View Slide

59. View Slide

60. View Slide

61. View Slide

62. View Slide

63. View Slide

64. View Slide

65. View Slide

66. View Slide

67. View Slide

68. View Slide

69. View Slide

70. View Slide

71. billing
    

    View Slide

72. THREE
    Embracing Failure
    

    View Slide

73. View Slide

74. View Slide

75. View Slide

76. View Slide

77. View Slide

78. “Do not go gentle into that good night”
    - Dylan Thomas
    

    View Slide

79. [reliable, distributed systems talk]
    

    View Slide

80. View Slide

81. If we can achieve reliable data,
    disposable computation comes easy
    

    View Slide

82. reliable data
    

    View Slide

83. lots of reliable data solutions
    

    View Slide

84. provide an atomic unit
    

    View Slide

85. easily addressable and discoverable
    

    View Slide

86. disposable computation
    

    View Slide

87. stateless
    

    View Slide

88. View Slide

89. View Slide

90. manage as batches
    

    View Slide

91. manage as commits
    

    View Slide

92. manage as transactions
    

    View Slide

93. manage as batches
    

    View Slide

94. batches are atomic
    

    View Slide

95. no state held between batches
    

    View Slide

96. independent
    

    View Slide

97. independent
    

    View Slide

98. restart at any batch point
    

    View Slide

99. crash and burn
    

    View Slide

100. clean restart
     

     View Slide

101. batch size matters
     

     View Slide

102. View Slide

103. services not servers
     

     View Slide

104. View Slide

105. View Slide

106. View Slide

107. View Slide

108. View Slide

109. View Slide

110. View Slide

111. respond to change
     

     View Slide

112. curl http://169.254.169.254/latest/meta-data/spot/termination-time
     

     View Slide

113. gcloud compute instances create example-instance \
     --metadata-from-file \
     shutdown-script=examples/scripts/install.sh
     

     View Slide

114. /etc/acpi/events/
     

     View Slide

115. reliable storage
     

     View Slide

116. disposable services
     

     View Slide

117. services backplane
     

     View Slide

118. failure i/o
     

     View Slide

119. FOUR
     Leverage Resilience
     

     View Slide

120. savings
     

     View Slide

121. peak spend
     

     View Slide

122. over-priced
     

     View Slide

123. under-
     priced
     

     View Slide

124. View Slide

125. excess capacity
     

     View Slide

126. savings
     

     View Slide

127. IT ALL COMES BACK TO SHIFTING
     COSTS
     

     View Slide

128. know what can be delayed
     

     View Slide

129. know what can be skipped
     

     View Slide

130. know what can be retried
     

     View Slide

131. sla
     

     View Slide

132. View Slide

133. price-aware deployment and scheduling
     

     View Slide

134. View Slide

135. budget & constraints
     

     View Slide

136. billing
     

     View Slide

137. sla
     

     View Slide

138. failure i/o
     

     View Slide

139. services backplane
     

     View Slide

140. View Slide

141. View Slide

142. View Slide

143. View Slide

144. View Slide

145. View Slide

146. the game of price
     

     View Slide

147. View Slide

148. intelligent
     co-ordination
     

     View Slide

149. FIVE
     Operational Simplicity
     

     View Slide

150. how much for new feature?
     

     View Slide

151. how much for new software?
     

     View Slide

152. how much for new environment?
     

     View Slide

153. how much for new location?
     

     View Slide

154. how much for new product?
     

     View Slide

155. flexibility and automation are core
     

     View Slide

156. but…
     

     View Slide

157. moving parts are costly
     

     View Slide

158. parts vs moving parts
     

     View Slide

159. split batch and online systems
     

     View Slide

160. split reliable and smart systems
     

     View Slide

161. HOW CAN OUR SOFTWARE ADAPT TO
     COST?
     

     View Slide

162. View Slide

163. budget & constraints
     

     View Slide

164. price-aware backplane
     

     View Slide

165. billing
     

     View Slide

166. reliable storage
     

     View Slide

167. disposable services
     

     View Slide

168. services backplane
     

     View Slide

169. sla
     

     View Slide

170. intelligent
     co-ordination
     

     View Slide

171. View Slide

172. G PRICE
     AWARE
     ARCHITECTURE
     

     View Slide