LTKA-05

Layanan Tersambung dan Komputasi Awan (Connected Services and Cloud Computing) - Cloud Computing

1. ET-3010 CLOUD COMPUTING (KOMPUTASI AWAN) H1/2014 Dr.-Ing. Eueung Mulyana School of Electrical Engineering and Informatics Institut Teknologi Bandung http://eueung.github.io/et3010-ltka
2. OUTLINE 1. Komputasi Awan - Revisited 2. Peran - Revisited 3. Overview 4. Computing at Scale 5. Data Center 6. Cloud and Utility Computing 7. PAYG Models 8. Challenges
3. KOMPUTASI AWAN - REVISITED
4. MODEL DAN RELASI Connected Services dan Cloud
5. MODEL - ZOOM IN Clouds di dalam Cloud
6. KA (CLOUD COMPUTING - CC) The interesting thing about Cloud Computing is that we've redefined Cloud Computing to include everything that we already do.... I don't understand what we would do differently in the light of Cloud Computing other than change the wording of some of our ads. (Larry Ellison, WSJ, 2008) A lot of people are jumping on the [cloud] bandwagon, but I have not heard two people say the same thing about it. There are multiple definitions out there of "the cloud". (Andy Isherwood, ZDnet News, 2008)
7. KA/CC (dilbert.com)
8. KA/CC Cloud computing is a model for enabling convenient, on-demand network access to a shared pool of configurable computing resources (e.g., networks, servers, storage, applications, and services) that can be rapidly provisioned and released with minimal management effort or service provider interaction. (NIST)
9. PERAN - REVISITED
10. PERAN KA/CC Apa perlunya KA?
11. FACEBOOK (Mung Chiang, Coursera)
12. TWITTER (Mung Chiang, Coursera)
13. ANGKA-ANGKA | USERS & OBJECTS 1. Facebook : 1.15 B pengguna aktif 2. Google : 1.2+ B queries/day dengan 27 B items 3. YouTube : 2+ B videos/day 4. Flickr : 6+ B foto Haeberlen, Ives (Univ. of Pennsylvania, 2013)
14. ANGKA-ANGKA | DATA 1. Google processed 20 PB per day (2008) 2. Rendering 'Avatar' movie required 1+ PB of storage 3. eBay has 6.5+ PB of user data 4. CERN's LHC will produce about 15 PB of data per year 5. German Climate computing center dimensioned for 60 PB of climate data 6. Google now designing for 1 EB of storage 7. NSA Utah Data Center is said to have 5 ZB (ada rumor 1 YB) Haeberlen, Ives (Univ. of Pennsylvania, 2013)
15. ANGKA-ANGKA | COMPUTATION 1. Facebook is thought to have more than 60,000 servers 2. Intel has +/- 100,000 servers in 97 data centers 3. Microsoft reportedly had at least 200,000 servers (2008) 4. Akamai has 95,000 servers in 71 countries 5. Google is thought to have more than 1 million servers, is planning for 10 million (according to Jeff Dean) Haeberlen, Ives (Univ. of Pennsylvania, 2013)
16. OVERVIEW
17. INTEPRETATION Rebranding of Web 2.0 Utility Computing Everything as a Service Large-Data Processing
18. REBRANDING OF WEB 2.0 Rich, interactive web applications Clouds refer to the servers that run them AJAX as the de facto standard (for better or worse) Examples: Facebook, YouTube, Gmail, ... The network is the computer User data is stored in the clouds Rise of the tablets, smartphones, etc. Browser is the OS
19. UTILITY COMPUTING What? Computing resources as a metered service (pay as you go) Ability to dynamically provision virtual machines Why? Cost: capital vs. operating expenses Scalability: infinite capacity Elasticity: scale up or down on demand Does it make sense? Benefits to cloud users Business case for cloud providers
20. EVERYTHING AS A SERVICE Infrastructure as a Service (IaaS) - Utility Computing Why buy machines when you can rent them? Examples: Amazon's EC2, Rackspace Platform as a Service (PaaS) - Also UC Give me nice API and take care of the maintenance, upgrades, ... Example: Google App Engine Software as a Service (SaaS) Just run it for me! Example: Gmail, Salesforce
21. LARGE-DATA Google processed 20 PB per day (2008) Google now designing for 1 EB of storage Rendering 'Avatar' movie required 1+ PB of storage eBay has 6.5+ PB of user data eBay has 10 PB Hadoop/Teradata, 75 B DB-calls a day (6/2012) CERN's LHC will produce about 15 PB of data per year German Climate computing center dimensioned for 60 PB of climate data NSA Utah Data Center is said to have 5 ZB (rumoured 1 YB)
22. LARGE-DATA Facebook: 100+ PB of user data; +500 TB/day (8/2012) Internet Archive Wayback Machine: 10 PB web archive (10/2012) LSST: 6-10 PB a year (appx. 2015)
25. (pewresearch.org)
26. COMPUTING AT SCALE
27. The Need
28. COMPUTING AT SCALE Why? Why should I care? Scale of Current and Future Services Current Services: How many users and objects? How much data? How much computation?
29. HOW MANY USERS AND OBJECTS? 1. Facebook : 1.15 B active users (2013) 2. Google : 1.2+ B queries/day on more than 27 B items 3. YouTube : 2+ B videos/day 4. Flickr : 6+ B photos
30. HOW MUCH DATA? Google processed 20 PB per day (2008) Google now designing for 1 EB of storage Rendering 'Avatar' movie required 1+ PB of storage eBay has 6.5+ PB of user data eBay has 10 PB Hadoop/Teradata, 75 B DB-calls a day (6/2012) CERN's LHC will produce about 15 PB of data per year German Climate computing center dimensioned for 60 PB of climate data NSA Utah Data Center is said to have 5 ZB (rumoured 1 YB)
31. Stack of 1TB Hardisks for 1 ZB (Haeberlen) Approximately 2x Earth Diameter To process that much data ...
32. HOW MUCH COMPUTATION? 1. Facebook is thought to have more than 60,000 servers 2. Intel has +/- 100,000 servers in 97 data centers 3. Microsoft reportedly had at least 200,000 servers (2008) 4. Akamai has 95,000 servers in 71 countries 5. Google is thought to have more than 1 million servers, is planning for 10 million (according to Jeff Dean) 6. 1&1 Internet has over 70,000 servers
33. IMAGINE ... Suppose you want to build the next Google How do you... ... download and store billions of web pages and images? ... quickly find the pages that contain a given set of terms? ... find the pages that are most relevant to a given search? ... answer 1.2 billion queries of this type every day?
34. IMAGINE ... Suppose you want to build the next Facebook How do you... ... store the profiles of over 500 million users? ... avoid losing any of them? ... find out which users might want to be friends?
35. From PCs to Data Centers
36. SCALING UP What if one computer is not enough? Buy a bigger (server- class) computer What if the biggest computer is not enough? Buy many computers (cluster)
37. SCALING UP PC Server Cluster (Haeberlen)
38. SCALING UP - CLUSTER Cluster Switch Nodes/Blades Storage (Haeberlen)
39. SCALING UP What if your cluster is too big (hot, power hungry) to fit into your office building? Build a separate building for the cluster Building can have lots of cooling and power Result: Data Center
40. SCALING UP PC Server Cluster Data Center (Haeberlen)
41. SCALING UP What if even a data center is not big enough? Build additional data centers Where? How many?
42. SCALING UP PC Server Cluster Data Center Data Centers Network of
43. EXAMPLE Global DC Network - Google
44. Problems ...
45. CLASSICAL SCALING TECHNIQUES P#1: Difficult to Dimension P#2: Expensive P#3: Difficult to Scale
46. P#1: DIFFICULT TO DIMENSION Load can vary considerably Peak load can exceed average load by factor 2x-10x But: Few users deliberately provision for less than the peak Result: Server utilization in existing data centers 5%-20%!! Dilemma: Waste resources or lose customers!
47. Provisioning for the peak load
48. Provisioning below the peak
49. P#2: EXPENSIVE Need to invest many $$$ in hardware Even a small cluster can easily cost $100,000 Microsoft recently invested $499 million in a single data center Need expertise Planning and setting up a large cluster is highly nontrivial Cluster may require special software, etc. Need maintenance Someone needs to replace faulty hardware, install software upgrades, maintain user accounts, etc.
50. P#3: DIFFICULT TO SCALE Scaling up is difficult Need to order new machines, install them, integrate with existing cluster - can take weeks Large scaling factors may require major redesign, e.g., new storage system, new interconnect, new building (!) Scaling down is difficult What to do with superfluous hardware? Server idle power is about 60% of peak Energy is consumed even when no work is being done Many fixed costs, such as construction
51. Computation at Scale
52. SUMMARY Modern applications require huge amounts of processing and data Measured in PB, millions of users, billions of objects Need special hardware, algorithms, tools to work at this scale Clusters and data centers can provide the resources we need Main difference: Scale (room-sized vs. building-sized) Special hardware; power and cooling are big concerns Clusters and data centers are not perfect Difficult to dimension; expensive; difficult to scale
53. DATA CENTER
54. Previously ... (LTKA-02)
55. DATA CENTER Masih ingat e.g.: Google 1 M Server (2008) ? → Pusat Data Tempat Menyimpan dan Mengolah Data Fasilitas atau tempat khusus untuk menyimpan resource peralatan yg diperlukan untuk keberlangsungan layanan tersambung atau untuk fungsionalitas operasi IT perusahaan Komponen: Server, Storage, Network →
56. GOOGLE DATA CENTER 1:45 A Look into a Google Data Center
57. Clusters & Data Center
58. CLUSTERS Cluster Switch Nodes/Blades Storage (Haeberlen)
59. CLUSTERS Network switch (connects nodes with each other and with other racks) Many nodes/blades (often identical) Storage devices Characteristics of a cluster: Many similar machines, close interconnection (same room?) Often special, standardized hardware (racks, blades) Usually owned and used by a single organization
60. POWER AND COOLING Clusters need lots of power Example: 140 Watts per server Rack with 32 servers: 4.5kW (needs special power supply!) Most of this power is converted into heat Large clusters need massive cooling 4.5kW is about 3 space heaters (portable) And that's just one rack!
61. ENERGY & DATA CENTER
62. ENERGY & DATA CENTER Data centers consume a lot of energy Makes sense to build them near sources of cheap electricity Example: Price per KWh is 3.6ct in Idaho (near hydroelectric power), 10ct in California (long distance transmission), 18ct in Hawaii (must ship fuel) Most of this is converted into heat - Cooling is a big issue!
63. Google Data Center
65. Google Data Center Previously ... (LTKA-02)
70. CATATAN Fasilitas Data Center Global Google - Backbone in the Cloud
71. Google Data Center The Dalles, Oregon (2006)
76. WHAT DOES A DC LOOK LIKE? Main DC Building (size of a football field)
77. WHAT DOES A DC LOOK LIKE? Cooling Plant
78. Google Data Center Singapore ... (2013E)
79. GOOGLE@SG Jurong West, 2.45 hectares (acquired 09/2011), online 12/2013 Google's first urban, multi-story data center! Robot themed offices A reuse water system that will pump recycled water through the colorful tubes ...
83. Google Data Center Others ...
88. Facebook Data Center ...
92. Back to Data Center
93. DATA CENTER A warehouse-sized computer A single data center can easily contain 10,000 racks with 100 cores in each rack (1,000,000 cores total)
94. WHAT'S IN A DATA CENTER? Hundreds or Thousands of Racks
95. WHAT'S IN A DATA CENTER? Massive Networking
96. WHAT'S IN A DATA CENTER? Emergency Power Supplies
97. WHAT'S IN A DATA CENTER? Massive Cooling
98. BUILDING BLOCKS Server Rack Cluster (Barroso, Hoelzle via Lin)
99. STORAGE HIERARCHY (Barroso, Hoelzle via Lin)
101. Trends + Others ...
102. MODULAR DATA CENTER Container Data Center Need more capacity? Just deploy another container!
103. MODULAR DC (Haeberlen)
104. MODULAR DC (Haeberlen)
105. MODULAR DC (Haeberlen)
106. GLOBAL DISTRIBUTION Google's DC Network - (Haeberlen), Outdated
107. GLOBAL DISTRIBUTION Google's DC Network - (Hoelzle), Maybe Outdated
108. GLOBAL DISTRIBUTION Google's DC Locations - Early 2014
109. GLOBAL DISTRIBUTION Data centers are often globally distributed e.g. case above (Google) Why? Need to be close to users (physics!) Cheaper resources Protection against failures
110. A Look Inside US Data Centers WhoIsHostingThis Infographics 2014
124. CLOUD AND UTILITY COMPUTING
125. CLOUD COMPUTING - MICROSOFT 4:00 Microsoft - Journey to the Cloud
126. CLOUD COMPUTING - AWS 2:46 Amazon - AWS
127. This part is mainly based on the slides by Haeberlen and Ives
128. POWER PLANT ANALOGY Waterwheel
129. POWER PLANT ANALOGY Steam Engine
130. POWER PLANT ANALOGY It used to be that everyone had their own power source Challenges are similar to the cluster: Needs large up-front investment, expertise to operate, difficult to scale up/down...
131. SCALING - POWER PLANT ... Then people started to build large, centralized power plants with very large capacity...
132. SCALING - POWER PLANT Solar PP
133. SCALING - POWER PLANT Hydro/Water
134. SCALING - POWER PLANT Coal (?)
135. SCALING - POWER PLANT Geothermal (?)
136. METERED USAGE MODEL Power plants are connected to customers by a network Usage is metered, and everyone (basically) pays only for what they actually use
137. METERED USAGE MODEL Power Source Network Meter Customer From Source to Customer
138. WHY IS THIS A GOOD THING? - ELECTRICITY Economies of scale: Cheaper to run one big power plant than many small ones Statistical multiplexing: High utilization! No up-front commitment: No investment in generator; pay-as- you-go model Scalability: Thousands of kilowatts available on demand; add more within seconds
139. WHY IS THIS A GOOD THING? - COMPUTING Economies of scale: Cheaper to run one big data center than many small ones Statistical multiplexing: High utilization! No up-front commitment: No investment in data center; pay- as-you-go model Scalability: Thousands of computers available on demand; add more within seconds
140. CLOUD COMPUTING - AGAIN Cloud computing is a model for enabling convenient, on-demand network access to a shared pool of configurable computing resources (e.g., networks, servers, storage, applications, and services) that can be rapidly provisioned and released with minimal management effort or service provider interaction. (NIST)
141. ESSENTIAL CHARACTERISTICS - CC On-demand self service Broad network access Resource pooling Rapid elasticity Measured service
142. UTILITY COMPUTING - REVISITED What? Computing resources as a metered service (pay as you go) Ability to dynamically provision virtual machines Why? Cost: capital vs. operating expenses Scalability: infinite capacity Elasticity: scale up or down on demand Does it make sense? Benefits to cloud users Business case for cloud providers
143. RELATED TERMS Utility Computing The service being sold by a cloud Focuses on the business model (pay-as-you-go), similar to classical utility companies The Web The Internet's information sharing model Some web services run on clouds, but not all The Internet A network of networks Used by the web; connects (most) clouds to their customers
144. Cloud Types Based on Services
145. EVERYTHING AS A SERVICE - REVISITED Infrastructure as a Service (IaaS) - Utility Computing Why buy machines when you can rent them? Examples: Amazon's EC2, Rackspace Platform as a Service (PaaS) - Also UC Give me nice API and take care of the maintenance, upgrades, ... Example: Google App Engine Software as a Service (SaaS) Just run it for me! Example: Gmail, Salesforce
146. EVERYTHING AS A SERVICE What kind of service does the cloud provide? Does it offer an entire application, or just resources? If resources, what kind / level of abstraction? Three types commonly distinguished SaaS, PaaS, IaaS Other XaaS types have been defined, but are less common: Desktop, Backend, Communication, Network, Monitoring, ...
147. SAAS
148. SAAS Cloud provides an entire application (Word processor, spreadsheet, CRM software, calendar...) Customer pays cloud provider (SaaS) Example: Google Apps, Salesforce.com
149. PAAS
150. PAAS Cloud provides middleware/infrastructure (e.g. Microsoft Common Language Runtime - CLR) Customer pays SaaS provider for the service; SaaS provider pays the cloud (PaaS) provider for the infrastructure Example: Windows Azure, Google App Engine
151. IAAS
152. IAAS Cloud provides raw computing resources (Virtual machine, blade server, hard disk, ...) Customer pays SaaS provider for the service; SaaS provider pays the cloud (IaaS) provider for the resources Examples: Amazon Web Services, Rackspace Cloud, GoGrid
153. Cloud Types Based on Access
154. CLOUDS BASED ON ACCESS
155. WHO CAN BECOME A CUSTOMER OF THE CLOUD? Public cloud: Commercial service; open to (almost) anyone. Example: Amazon AWS, Microsoft Azure, Google App Engine Community cloud: Shared by several similar organizations.Example: Google's Gov Cloud Private cloud: Shared within a single organization. Example: Internal datacenter of a large company. Hybrid cloud: Private + Public
156. Enterprise Clouds iWeb Infographics 2013 Public, Private & Hybrid Clouds + Some Numbers ...
164. Cases ...
165. ANIMOTO Lets users create videos from their own photos/music Auto-edits photos and aligns them with the music, so it looks good Built using Amazon EC2+S3+SQS Released a Facebook app in mid-April 2008 More than 750,000 people signed up within 3 days EC2 usage went from 50 machines to 3,500 (x70 scalability!)
166. THE WASHINGTON POST March 19, 2008: Hillary Clinton's official White House schedule released to the public 17,481 pages of non-searchable, low-quality PDF Very interesting to journalists, but would have required hundreds of man-hours to evaluate Peter Harkins, Senior Engineer at The Washington Post: Can we make that data available more quickly, ideally within the same news cycle? Tested various Optical Character Recognition (OCR) programs; estimated required speed Launched 200 EC2 instances; project was completed within nine hours (!) using 1,407 hours of VM time ($144.62) Results available on the web only 26 hours after the release
167. OTHERS DreamWorks is using the Cerelink cloud to render animation movies (Cloud was already used to render parts of Shrek Forever After and How to Train your Dragon) CERN is working on a science cloud to process experimental data Virgin Atlantic is hosting their new travel portal on Amazon AWS
168. Cloud and Utility Computing
169. SUMMARY Why is cloud computing attractive? Analogy to 'classical' utilities (electricity, water, ...) No up-front investment (pay-as-you-go model) Low price due to economies of scale Elasticity - can quickly scale up/down as demand varies Different types of clouds SaaS, PaaS, IaaS; public/community/private/hybrid clouds
170. SUMMARY What runs on the cloud? Many potential applications: Application hosting, backup/storage, scientific computing, content delivery, ... Not yet suitable for certain applications (sensitive data, compliance requirements)
171. IS THE CLOUD GOOD FOR EVERYTHING? No Sometimes it is problematic, e.g., because of auditability requirements Examples: Processing medical records, Processing financial information Besides, would you put your medical data on a (public) cloud?
172. SUMMARY Clouds are good for many things... Applications that involve large amounts of computation, storage, bandwidth Especially when lots of resources are needed quickly (Washington Post example) or load varies rapidly ... but not for all things
173. GOOGLE CLOUD PLATFORM 3:00 DevBytes - Google Cloud Platform
174. PAYG MODELS
175. PAYG MODELS Pay-As-You-Go Models or Pay-As-You-Grow Models or simply be refered to as the Cloud Models
176. Amazon Web Services
183. Google Cloud Platform
186. Heroku PaaS
190. CHALLENGES
191. This part is mainly based on the slides by Haeberlen and Ives
192. 10 Obstacles & Opportunities
193. CHALLENGES Availability What happens to my business if there is an outage in the cloud? Data Lock-In How do I move my data from one cloud to another? Data Confidentiality and Auditability How do I make sure that the cloud doesn't leak my confidential data? Can I comply with regulations (e.g. HIPAA and Sarbanes/Oxley)?
194. OUTAGE & PERFORMANCE
195. CHALLENGES Data Transfer Bottlenecks How do I copy large amounts of data from/to the cloud? Example: 10 TB from UC Berkeley to Amazon in Seattle, WA (Internet 20 Mbps: 45 days; FedEx 1 day) Motivated Import/Export feature on AWS Performance Unpredictability Example: VMs sharing the same disk - I/O interference Example: HPC tasks that require coordinated scheduling
196. CHALLENGES Scalable Storage Cloud model (short-term usage, no up-front cost, infinite capacity on demand) does not fit persistent storage well Bugs in Large Distributed Systems Many errors cannot be reproduced in smaller configs Scaling Quickly Problem: Boot time; idle power Fine-grain accounting?
197. CHALLENGES Reputation Fate Sharing One customer's bad behavior can affect the reputation of others using the same cloud Example: Spam blacklisting, FBI raid after criminal activity Software Licensing What if licenses are for specific computers? (e.g. Microsoft Windows) How to scale number of licenses up/down? (Need pay-as- you-go model as well)
198. CREDITS
199. CREDITS 1. Jimmy Lin, Cloud Computing and Big Data, The iSchool, Univ. of Maryland, 2013 2. A. Haeberlen, Z. Ives, Scalable and Cloud Computing, Univ. of Pennsylvania, 2013