TDC Floripa 2016 - Cloud Computing

Transcript

1. ©  2015,  Amazon  Web  Services,  Inc.  or  its  Affiliates.  All  rights  reserved. Julio  M.  Faerman @jmfaerman The  Developer’s  Conference Florianópolis 2016 Inovação na Escala da  AWS (SPOT301)  AWS  Innovation  at  Scale
2. Princípios Operacionais
3. Capacidade sob  demanda
4. Confiabilidade Capacidade sob  demanda
5. Confiabilidade Baixo custo Capacidade sob  demanda
6. Escala
7. Escala Melhor Capacidade
8. Mais Robustez Melhor Capacidade Escala
9. Menor custo Mais Robustez Melhor Capacidade Escala
10. Not  that  Long  Ago…
11. Not  that  Long  Ago… • 11-­15%  utilization
12. Not  that  Long  Ago… • 19%  utilization
13. Not  that  Long  Ago… • 19%  utilization • 81%  idle
14. Not  that  Long  Ago… • 19%  utilization • 81%  idle • Lots  of  meetings
15. Not  that  Long  Ago… • 19%  utilization • 81%  idle • Lots  of  meetings • 3  to  9  months  to   deploy  apps
16. Partial  reliability
17. Partial  reliability Unnecessary  and   missing  features
18. Partial  reliability Unnecessary  and   missing  features Outages
19. Poor  Utilization
20. Capacity  not   Available
21. Vicious  Cycle
22. O  que muda com  a  escala?
23. Land
24. Power
25. Power
26. Power  at  Scale
27. Power  Infrastructure • Power  generation  facilities • Negotiated  PPAs • Commitment  to  renewable   energy  sources
28. Amazon  Wind  Farm  (Fowler  Ridge) • Indiana • January  2015 • Pattern  Development • 150  MW  wind
30. Amazon  Wind  Farm  US  East • North  Carolina • July  2015 • 208  MW  wind
31. November 2014 Announce  commitment  to  100%  renewable  energy April  2015 Achieved  25% renewable  milestone Announced  4.8  MWh battery  pilot  with  Tesla April  2015 Joined  the  American  Council  on  Renewable  Energy   (ACORE) April  2015 Announced participation  in  the  U.S.  Partnership  for   Renewable  Energy  Finance End  of  2016 Goal  of  40%  renewable  energy  for  power  consumed AWS  Renewable  Energy  Commitment
32. Amazon  Solar  Farm  US  East • East  Virginia • June  2015 • Community  Energy • 80  MW  solar
33. Por que isso é mais eficiente?
34. Cloud  Efficiency Average  AWS  Customer… • Uses  77%  fewer  servers • Uses  84%  less  power
35. Cloud  Efficiency NRDC  2014  assessment… • Cloud  companies  at  65%  utilization
36. Cloud  Efficiency NRDC  2014  assessment… • Cloud  companies  at  65%  utilization • On  premises  at  ~15%  utilization
37. Cloud  Efficiency NRDC  2014  assessment… • Cloud  companies  at  65%  utilization • On  premises  at  ~15%  utilization • 15  ÷ 65  =  23%
38. Cloud  Efficiency Annual  Uptime  Institute  survey…
39. Cloud  Efficiency Annual  Uptime  Institute  survey… • PUE  of  1.7  =  enterprise  datacenter
40. Cloud  Efficiency Annual  Uptime  Institute  survey… • PUE  of  1.7  =  enterprise  datacenter • PUE  of  1.07  to  1.15  =  cloud  datacenter
41. Cloud  Efficiency Annual  Uptime  Institute  survey… • PUE  of  1.7  =  enterprise  datacenter • PUE  of  1.07  to  1.15  =  cloud  datacenter • 1.2  ÷ 1.7  =  71%   • A  29%  reduction  in  power  pulled  from  the  grid
42. Cloud  Efficiency What  this  means… • 23%  x  71%  =  16% • 84%  reduction  in  power
43. Customer  Benefit • 60  million  members • 10  billion  streaming  hours • ~  50%  renewable • Server  utilization  >50% • Reduced  server  costs  85%
44. Planning,  Design  &  Construction
45. Security
46. Security
47. Security
48. Security
49. Security
50. Security
51. Redes
52. Old  World Why  it  doesn’t  work… • Complex  hardware  /  software • Burdened  with  unnecessary  features   • Buggy,  slow  to  fix,  slow  to  improve • Complexity  equals  cost
53. Need  for  a  Purpose  Build  Network Advantages  of  purpose-­built  networks: • Simple  designs  at  each  layer • Hardware  &  software  not  burdened   with  unnecessary  features   • Simple  software  = less  hardware • Simplicity  =  lower  cost
54. AZ AZ AZ AZ AZ Purpose  Built  Network • Each  datacenter  has  a  purpose  built   network  
55. AZ AZ AZ AZ AZ Purpose  Built  Network
56. AZ AZ AZ AZ AZ Purpose  Built  Network
57. AZ AZ AZ AZ AZ Purpose  Built  Network • Metro-­area  DWDM  links  between  AZs • 82,864  fiber  strands  in  region • AZs  <  2ms  apart  &  usually  <1ms • 25+  Tbps inter-­AZs  peak  capacity • Each  datacenter  has  a  purpose  built   network  
58. AZ AZ AZ AZ AZ Transit Transit • Redundant  paths  to  transit  centers • Transit  Centers  connect  to: • Private  links  to  other  AWS  regions • Private  links  to  AWS  Direct  Connect  customers • Internet  through  peering  &  paid  transit Purpose  Build  Network • Metro-­area  DWDM  links  between  AZs • 82,864  fiber  strands  in  region • AZs  <2ms  apart  &  usually  <1ms • 25+  Tbps inter-­AZs  peak  capacity • Each  datacenter  has  a  purpose  built   network  
59. AWS  Worldwide  Network  Backbone • Provides  full,  direct  control • Reduces  cost,  complexity,  jitter • Increases  performance  &  availability • Better  Quality  of  Service  (QoS) • Not  over-­subscribed  and  fully  alarmed
60. Servers
61. Need  for  Customization Advantages  of  purpose-­built  servers: • Work  directly  with  component   manufacturers  like  memory,  HDD,  CPU • Design  optimized  for  specific  workloads • Hardware  &  software  not  burdened  with   unnecessary  features • Highly  utilized
62. Custom  Storage  Rack • 864  disks • 1065  Kg
63. Importance  of  Supply  Chain
64. Importance  of  Supply  Chain
65. Importance  of  Supply  Chain
66. Importance  of  Supply  Chain
67. Importance  of  Supply  Chain
68. Importance  of  Supply  Chain
69. Importance  of  Supply  Chain
70. Retail  Leverage
71. AWS  Worldwide  Locations India Ohio   Ningxia Montreal UK
72. *  As  of  1  Mar  2016 2009 48 280 722 82 2011 2013 2015 Ritmo de  Inovação Acelerado
73. http://blog.concretesolutions.com.br/2015/04/amazon-­machine-­learning-­benchmark/ http://blog.concretesolutions.com.br/2015/04/amazon-­machine-­learning-­benchmark/
77. Announcing  AWS  IoT AWS  IoT is  a  managed  cloud   platform  that  lets  connected  devices   easily  and  securely  interact  with   cloud  applications  and  other   devices.  AWS  IoT can  support   billions  of  devices  and  trillions  of   messages,  and  can  process  and   route  those  messages  to  AWS   endpoints,  such  as  AWS  Lambda   and  Amazon  Kinesis,  and  to  other   devices  reliably  and  securely. AWS  Mobile  Hub AWS  Mobile  Hub  is  the  fastest  way   to  build  mobile  apps  powered  by   AWS.  It  lets  you  easily  add  and   configure  features  for  your  apps,   including  user  authentication,  data   storage,  backend  logic,  push   notifications,  content  delivery,  and   analytics  — all  from  a  single,   integrated  console,  and  without   worrying  about  provisioning,  scaling,   and  managing  the  infrastructure.   AWS  Lambda   You  can  now  develop  your  AWS   Lambda  function  code  using  Python,   maintain  multiple  versions  of  your   function  code,  invoke  your  code  on   a  regular  schedule,  and  run  your   functions  for  up  to  five  minutes.  
78. AWS  Inspector Analyze  application  security Amazon  Inspector  is  an  automated   security  assessment  service  that  helps   minimize   the  likelihood   of  introducing   security  or  compliance  issues  when   deploying  applications   on  AWS.   Amazon  Inspector  automatically   assesses  applications  for  vulnerabilities   or  deviations  from  best  practices.  After   performing  an  assessment,  Amazon   Inspector  produces  a  detailed  report   with  prioritized  steps  for  remediation.   AWS  Database  Migration   Service Migrate  databases  to  AWS  with  minimal   downtime AWS  Database  Migration  Service  helps  you   migrate  databases  to  AWS  easily  and  securely   with  minimal  downtime.  It  supports  all  widely  used   database  platforms,  and  performs  schema  and   code  conversion  for  migrations  between  database   engines.  AWS  Database  Migration  Service  is  low   cost  and  simple  to  use.  You  only  pay  for  the   compute  resources  used  during  the  migration   process  and  can  typically  migrate  a  terabyte-­size   database  for  as  little  as  three  dollars.   Amazon  Kinesis  Firehose Easily  load  streaming  data  into  S3  &   Redshift Amazon  Kinesis  Firehose is  the  easiest   way  to  load  streaming  data  into  AWS.  It   can  capture  and  automatically  load   streaming  data  into  Amazon  S3  and   Amazon  Redshift,  enabling   near  real-­ time  analytics  with  existing  business   intelligence   tools  you’re  already  using   today.  It  is  a  fully  managed  service  that   requires  no  ongoing  administration.
79. Em resumo • Escala e  ritmo de  inovação sem precedentes • Demanda personalização em todas camadas • Inovação em todas dimensões de  um   datacenter • O  valor  entregue aos cliente se  extende além do  software
82. Thank  You Obrigado! Perguntas? Julio  M.  Faerman @jmfaerman