Cloud Computing. Ennan Zhai. Computer Science at Yale University

Transcription

1 Cloud Computing Ennan Zhai Computer Science at Yale University

2 About Final Project

3 About Final Project Important dates before demo session: - Oct 31: Proposal v1.0 - Nov 7: Source code v1.0 - Nov 14: Proposal v2.0 - Nov 21: Source code v2.0 - Dec 1: Bakeoff version

4 How to present? About Final Project - ~2 min video and ~2 min Q&A What to submit? - 1) Source code - 2) Proposal - 3) README Others - any programming language - running on zoo machines - individual or group (2-3 members)

5 About Final Project Topics: - Structured P2P system, e.g., Chord - Hybrid P2P system, e.g., BubbleStorm - client software - Decentralized read/write file system - Decentralized social network application - Sybil-resistant recommendation system - Fault tolerance (BFT) system - Accountability system, e.g., PeerReview

6 About Final Project A few good final projects in the past years: - Workable client - A picture-based encryption tool - A privacy-preserving social network

7 Questions?

8 Lecture Outline Cloud Computing Challenges in the Clouds A Concrete Cloud Reliability Case

9 Lecture Outline Cloud Computing Challenges in the Clouds A Concrete Cloud Reliability Case

10 What s the Cloud Computing?

11 What s the Cloud Computing? Cloud computing is a business model for enabling convenient network access to a shared pool of configurable resources which can be rapidly provisioned and released with minimal management effort or service provider interaction. --- according to NIST(National Institute of Standards and Technology)

12 Have You Used the Cloud?

13 Have You Used the Cloud?

14 Have You Used the Cloud?

15 Have You Used the Cloud?

16 Have You Used the Cloud?

17 Have You Used the Cloud?

18 Why We Like It?

19 Why We Like It? Why users like it? - Do not care where it is, it is just there - Access from any platform Why CS researchers like it? - High-performance computation with less money - Lots of hard and interesting challenges

20 Why We Like It? Why users like it? - Do not care where it is, it is just there - Access from any platform Why CS researchers like it? - High-performance computation with less money - Lots of hard and interesting challenges

21 Why We Like It? Why users like it? - Do not care where it is, it is just there - Access from any platform Why CS researchers like it? - High-performance computation with less money - Lots of hard and interesting challenges

22 Cloud V.S. Distributed Systems

23 What Kinds of Clouds Exist Now?

24 What Kinds of Clouds Exist Now? Three types of services: - Software as a Service (SaaS) - Analogy: Restaurant. Prepares&serves entire meal, does the dishes, etc - Platform as a Service (PaaS) - Analogy: Take-out food. Prepares meal but does not serve it. - Infrastructure as a Service (IaaS) - Analogy: Grocery store. Provides raw ingredients.

25 What Kinds of Clouds Exist Now? Three types of services: - Software as a Service (SaaS) - Analogy: Restaurant. Prepares&serves entire meal, does the dishes, etc - Platform as a Service (PaaS) - Analogy: Take-out food. Prepares meal but does not serve it. - Infrastructure as a Service (IaaS) - Analogy: Grocery store. Provides raw ingredients.

26 What Kinds of Clouds Exist Now? Three types of services: - Software as a Service (SaaS) - Analogy: Restaurant. Prepares&serves entire meal, does the dishes, etc - Platform as a Service (PaaS) - Analogy: Take-out food. Prepares meal but does not serve it. - Infrastructure as a Service (IaaS) - Analogy: Grocery store. Provides raw ingredients.

27 What Kinds of Clouds Exist Now? Three types of services: - Software as a Service (SaaS) - Analogy: Restaurant. Prepares&serves entire meal, does the dishes, etc - Platform as a Service (PaaS) - Analogy: Take-out food. Prepares meal but does not serve it. - Infrastructure as a Service (IaaS) - Analogy: Grocery store. Provides raw ingredients.

28 Software as a Service (SaaS)

29 Software as a Service (SaaS) Cloud Provider (i.e., SaaS Provider) Application Middleware Hardware

30 Software as a Service (SaaS) Cloud Provider (i.e., SaaS Provider) Application Middleware Hardware SaaS provider offers an entire application - Word processor, spreadsheet, CRM software, etc. - Customer pays cloud provider - Example: Google Apps, Salesforce.com, etc.

31 Software as a Service (SaaS) Cloud Provider (i.e., SaaS Provider) Application Middleware Hardware SaaS provider offers an entire application - Word processor, spreadsheet, CRM software, etc. - Customer pays cloud provider - Example: Google Apps, Salesforce.com, etc.

32 Software as a Service (SaaS) Customer Cloud Provider (i.e., SaaS Provider) Application Middleware Hardware SaaS provider offers an entire application - Word processor, spreadsheet, CRM software, etc. - Customer pays cloud provider and uses the service - Example: Google Apps, Salesforce.com, etc.

33 Software as a Service (SaaS) Customer Cloud Provider (i.e., SaaS Provider) Application Middleware Hardware SaaS provider offers an entire application - Word processor, spreadsheet, CRM software, etc. - Customer pays cloud provider and uses the service - Example: Google Apps, Salesforce.com, etc.

34 Software as a Service (SaaS) Customer Cloud Provider (i.e., SaaS Provider) Application Middleware Hardware SaaS provider offers an entire application - Word processor, spreadsheet, CRM software, etc. - Customer pays cloud provider and uses the service - Example: Google Apps, Salesforce.com, etc.

35 A Typical SaaS: Gmail

36 A Typical SaaS: Gmail Gmail Provider Application Middleware Hardware

37 A Typical SaaS: Gmail Gmail Provider Application Middleware Hardware Outsourcing your software: - Distributed, replicated message store in BigTable - Weak consistency model for some operations (e.g., msg read) - Stronger consistency for others (e.g., send msg)

38 A Typical SaaS: Gmail Gmail Provider Application Middleware Hardware BigTable Outsourcing your software: - Distributed, replicated message store in BigTable - Weak consistency model for some operations (e.g., msg read) - Stronger consistency for others (e.g., send msg)

39 A Typical SaaS: Gmail Gmail Provider Application BigTable Middleware APIs Hardware BigTable Outsourcing your software: - Distributed, replicated message store in BigTable - Weak consistency model for some operations (e.g., msg read) - Stronger consistency for others (e.g., send msg)

40 A Typical SaaS: Gmail Gmail Provider Application Gmail BigTable Middleware APIs Hardware BigTable Outsourcing your software: - Distributed, replicated message store in BigTable - Weak consistency model for some operations (e.g., msg read) - Stronger consistency for others (e.g., send msg)

41 A Typical SaaS: Gmail Customer Gmail Provider Application Gmail BigTable Middleware APIs Hardware BigTable Outsourcing your software: - Distributed, replicated message store in BigTable - Weak consistency model for some operations (e.g., msg read) - Stronger consistency for others (e.g., send msg)

42 A Typical SaaS: Gmail Customer Gmail Provider Application Gmail BigTable Middleware APIs Hardware BigTable Outsourcing your software: - Distributed, replicated message store in BigTable - Weak consistency model for some operations (e.g., msg read) - Stronger consistency for others (e.g., send msg)

43 A Typical SaaS: Gmail Customer Gmail Provider Application Gmail BigTable Middleware APIs Hardware BigTable Outsourcing your software: - Distributed, replicated message store in BigTable - Weak consistency model for some operations (e.g., msg read) - Stronger consistency for others (e.g., send msg)

44 A Typical SaaS: Gmail Customer Gmail Provider Application Gmail BigTable Middleware APIs Hardware BigTable Outsourcing your software: - Distributed, replicated message store in BigTable - Weak consistency model for some operations (e.g., msg read) - Stronger consistency for others (e.g., send msg)

45 Platform as a Service (PaaS)

46 Platform as a Service (PaaS) Cloud Provider (i.e., PaaS Provider) Application Middleware Hardware Cloud provides middleware/infrastructure - For example, Microsoft Common Language Runtime (CLR) - Customer pays SaaS provider for the service - SaaS provider pays the cloud for the platform - Example: Windows Azure, Google App Engine, etc.

47 Platform as a Service (PaaS) Cloud Provider (i.e., PaaS Provider) Application Middleware Hardware Cloud provides middleware/infrastructure - For example, Microsoft Common Language Runtime (CLR) - Customer pays SaaS provider for the service - SaaS provider pays the cloud for the platform - Example: Windows Azure, Google App Engine, etc.

48 Platform as a Service (PaaS) App Provider Cloud Provider (i.e., PaaS Provider) Application Middleware Hardware Cloud provides middleware/infrastructure - For example, Microsoft Common Language Runtime (CLR) - App provider pays the cloud for the platform - Customer pays App provider for the service - Example: Windows Azure, Google App Engine, etc.

49 Platform as a Service (PaaS) App Provider Cloud Provider (i.e., PaaS Provider) Application Middleware Hardware Cloud provides middleware/infrastructure - For example, Microsoft Common Language Runtime (CLR) - App provider pays the cloud for the platform - Customer pays App provider for the service - Example: Windows Azure, Google App Engine, etc.

50 Platform as a Service (PaaS) App Provider Customer Application Middleware Hardware Cloud Provider (i.e., PaaS Provider) Cloud provides middleware/infrastructure - For example, Microsoft Common Language Runtime (CLR) - App provider pays the cloud for the platform - Customer pays app provider for the service - Example: Windows Azure, Google App Engine, etc.

51 Platform as a Service (PaaS) App Provider Customer Application Middleware Hardware Cloud Provider (i.e., PaaS Provider) Cloud provides middleware/infrastructure - For example, Microsoft Common Language Runtime (CLR) - App provider pays the cloud for the platform - Customer pays app provider for the service - Example: Windows Azure, Google App Engine, etc.

52 Platform as a Service (PaaS) App Provider Customer Application Middleware Hardware Cloud Provider (i.e., PaaS Provider) Cloud provides middleware/infrastructure - For example, Microsoft Common Language Runtime (CLR) - App provider pays the cloud for the platform - Customer pays app provider for the service - Example: Windows Azure, Google App Engine, etc.

53 A Typical PaaS: Facebook

54 A Typical PaaS: Facebook Application Middleware Facebook Provider Hardware

55 A Typical PaaS: Facebook Facebook Provider Application Middleware Hardware Facebook offers PaaS capabilities to App provider - Facebook APIs allow access to social network properties - Third-party game applications - Facebook itself also uses PaaS provided by its company, e.g., log analysis for recommendations

56 A Typical PaaS: Facebook Facebook Provider Application Facebook Middleware APIs Facebook Hardware Clusters Facebook offers PaaS capabilities to App provider - Facebook APIs allow access to social network properties - Third-party game applications - Facebook itself also uses PaaS provided by its company, e.g., log analysis for recommendations

57 A Typical PaaS: Facebook Facebook Provider Application Facebook Middleware APIs Facebook Hardware Clusters Facebook offers PaaS capabilities to App provider - Facebook APIs allow access to social network properties - Third-party game applications - Facebook itself also uses PaaS provided by its company, e.g., log analysis for recommendations

58 A Typical PaaS: Facebook App Provider Facebook Application Game Facebook Middleware APIs Facebook Hardware Clusters Facebook Provider Facebook offers PaaS capabilities to App provider - Facebook APIs allow access to social network properties - App providers adopt their services (e.g., game) onto Facebook - Facebook itself also uses PaaS provided by its company, e.g., log analysis for recommendations

59 A Typical PaaS: Facebook App Provider Facebook Application Game Facebook Middleware APIs Facebook Hardware Clusters Facebook Provider Facebook offers PaaS capabilities to App provider - Facebook APIs allow access to social network properties - App providers adopt their services (e.g., game) onto Facebook - Facebook itself also uses PaaS provided by its company, e.g., log analysis for recommendations

60 A Typical PaaS: Facebook App Provider Customer Facebook Application Game Facebook Middleware APIs Facebook Hardware Clusters Facebook Provider Facebook offers PaaS capabilities to App provider - Facebook APIs allow access to social network properties - App providers adopt their services (e.g., game) onto Facebook - Facebook itself also uses PaaS provided by its company, e.g., log analysis for recommendations

61 A Typical PaaS: Facebook App Provider Customer Facebook Application Game Facebook Middleware APIs Facebook Hardware Clusters Facebook Provider Facebook offers PaaS capabilities to App provider - Facebook APIs allow access to social network properties - App providers adopt their services (e.g., game) onto Facebook - Facebook itself also uses PaaS provided by its company, e.g., log analysis for recommendations

62 A Typical PaaS: Facebook App Provider Customer Facebook Application Game Facebook Middleware APIs Facebook Hardware Clusters Facebook Provider Facebook offers PaaS capabilities to App provider - Facebook APIs allow access to social network properties - App providers adopt their services (e.g., game) onto Facebook - Facebook itself also uses PaaS provided by its company, e.g., log analysis for recommendations

63 Infrastructure as a Service (IaaS)

64 Infrastructure as a Service (IaaS) Cloud Provider (i.e., IaaS Provider) Application Middleware Hardware Cloud provides raw computing resources - Virtual machines, blade servers, hard disk, etc. - Customer pays SaaS provider for the service - SaaS provider pays the cloud for the resources - Example: Amazon Web Services, Rackspace Cloud, etc.

65 Infrastructure as a Service (IaaS) Cloud Provider (i.e., IaaS Provider) Application Middleware Hardware Cloud provides raw computing resources - Virtual machines, blade servers, hard disk, etc. - Customer pays SaaS provider for the service - SaaS provider pays the cloud for the resources - Example: Amazon Web Services, Rackspace Cloud, etc.

66 Infrastructure as a Service (IaaS) App Provider Cloud Provider (i.e., IaaS Provider) Application Middleware Hardware Cloud provides raw computing resources - Virtual machines, blade servers, hard disk, etc. - App provider pays the cloud for the resources - Customer pays App provider for the service - Example: Amazon Web Services, Rackspace Cloud, etc.

67 Infrastructure as a Service (IaaS) App Provider Cloud Provider (i.e., IaaS Provider) Application Middleware Hardware Cloud provides raw computing resources - Virtual machines, blade servers, hard disk, etc. - App provider pays the cloud for the resources - Customer pays App provider for the service - Example: Amazon Web Services, Rackspace Cloud, etc.

68 Infrastructure as a Service (IaaS) App Provider Customer Application Middleware Hardware Cloud Provider (i.e., IaaS Provider) Cloud provides raw computing resources - Virtual machines, blade servers, hard disk, etc. - App provider pays the cloud for the resources - Customer pays App provider for the service - Example: Amazon Web Services, Rackspace Cloud, etc.

69 Infrastructure as a Service (IaaS) App Provider Customer Application Middleware Hardware Cloud Provider (i.e., IaaS Provider) Cloud provides raw computing resources - Virtual machines, blade servers, hard disk, etc. - App provider pays the cloud for the resources - Customer pays App provider for the service - Example: Amazon Web Services, Rackspace Cloud, etc.

70 Infrastructure as a Service (IaaS) App Provider Customer Application Middleware Hardware Cloud Provider (i.e., IaaS Provider) Cloud provides raw computing resources - Virtual machines, blade servers, hard disk, etc. - App provider pays the cloud for the resources - Customer pays App provider for the service - Example: Amazon Web Services, Rackspace Cloud, etc.

71 Typical IaaS: EC2 and S3 Application Middleware Amazon Hardware

72 Typical IaaS: EC2 and S3 Application Middleware Amazon EC2 Hardware S3

73 Typical IaaS: EC2 and S3 Netflix Provider Amazon Application Middleware EC2 Hardware S3 Netflix (app) heavily depends on Amazon AWS: - Media files are stored in S3 - Transcoding to target devices (e.g., ipad) using EC2 - Analysis of streaming sessions based on Elastic MapReduce

74 Typical IaaS: EC2 and S3 Netflix Provider Amazon Application Middleware EC2 Hardware S3 Netflix (app) heavily depends on Amazon AWS: - Media files are stored in S3 - Transcoding to target devices (e.g., ipad) using EC2 - Analysis of streaming sessions based on Elastic MapReduce

75 Typical IaaS: EC2 and S3 Netflix Provider Application Netflix Middleware Amazon EC2 Hardware S3 Netflix (app) heavily depends on Amazon AWS: - Media files are stored in S3 - Transcoding to target devices (e.g., ipad) using EC2 - Analysis of streaming sessions based on Elastic MapReduce

76 Typical IaaS: EC2 and S3 Netflix Provider Application Netflix Middleware Amazon EC2 Hardware S3 Netflix (app) heavily depends on Amazon AWS: - Media files are stored in S3 - Transcoding to target devices (e.g., ipad) using EC2 - Analysis of streaming sessions based on Elastic MapReduce

77 Typical IaaS: EC2 and S3 Netflix Provider Customer Application Netflix Middleware Amazon EC2 Hardware S3 Netflix (app) heavily depends on Amazon AWS: - Media files are stored in S3 - Transcoding to target devices (e.g., ipad) using EC2 - Analysis of streaming sessions based on Elastic MapReduce

78 Typical IaaS: EC2 and S3 Netflix Provider Customer Application Netflix Middleware Amazon EC2 Hardware S3 Netflix (app) heavily depends on Amazon AWS: - Media files are stored in S3 - Transcoding to target devices (e.g., ipad) using EC2 - Analysis of streaming sessions based on Elastic MapReduce

79 Recall Three types of services: - Software as a Service (SaaS) - Analogy: Restaurant. Prepares&serves entire meal, does the dishes, etc - Platform as a Service (PaaS) - Analogy: Take-out food. Prepares meal but does not serve it. - Infrastructure as a Service (IaaS) - Analogy: Grocery store. Provides raw ingredients.

80 Recall Three types of services: - Software as a Service (SaaS) - Analogy: Restaurant. Prepares&serves entire meal, does the dishes, etc Zoo? - Platform as a Service (PaaS) - Analogy: Take-out food. Prepares meal but does not serve it. - Infrastructure as a Service (IaaS) - Analogy: Grocery store. Provides raw ingredients.

81 The Major Cloud Providers

82 The Major Cloud Providers Amazon is the big player: - Infrastructure as a service (e.g., EC2) - Storage as a service (e.g., S3) But there are many others: - Microsoft Azure: It has similar services to Amazon, with an emphasis on.net programming model - Google App Engine: It offers programming interface, Hadoop, also software as a service, e.g., Gmail and Google Docs - IBM, HP, Yahoo!: They seem to focus on enterprise scale cloud apps

83 The Major Cloud Providers Amazon is the big player: - Infrastructure as a service (e.g., EC2) - Storage as a service (e.g., S3) But there are many others: - Microsoft Azure: It has similar services to Amazon, with an emphasis on.net programming model - Google App Engine: It offers programming interface, Hadoop, also software as a service, e.g., Gmail and Google Docs - IBM, HP, Yahoo!: They seem to focus on enterprise scale cloud apps

84 Lecture Outline Cloud Computing Challenges in the Clouds A Concrete Cloud Reliability Case

85 Lecture Outline Cloud Computing Challenges in the Clouds A Concrete Cloud Reliability Case

86 What Kinds of Challenges?

87 What Kinds of Challenges? Scalability Availability and reliability Security and privacy

88 What Kinds of Challenges? Scalability Availability and reliability Security and privacy

89 PC Scalability

90 Scalability PC What if one computer is not enough? - Buy a bigger (server-class) computer

91 Scalability PC Server What if one computer is not enough? - Buy a bigger (server-class) computer

92 Scalability PC Server What if one computer is not enough? - Buy a bigger (server-class) computer What if the biggest computer is not enough? - Buy many computers

93 Scalability PC Server Cluster What if one computer is not enough? - Buy a bigger (server-class) computer What if the biggest computer is not enough? - Buy many computers

94 Scalability

95 Rack Scalability

96 Rack Scalability Network switches (connects nodes with each other and with other racks)

97 Rack Scalability Network switches (connects nodes with each other and with other racks) Many nodes/blades (often identical)

98 Rack Scalability Network switches (connects nodes with each other and with other racks) Many nodes/blades (often identical) Storage device(s)

99 Scalability PC Server Cluster What if cluster is too big to fit into machine room? - Build a separate building for the cluster - Building can have lots of cooling and power - Result: Data center

100 Scalability PC Server Cluster What if cluster is too big to fit into machine room? - Build a separate building for the cluster - Building can have lots of cooling and power - Result: Data center

101 Scalability PC Server Cluster Data center What if cluster is too big to fit into machine room? - Build a separate building for the cluster - Building can have lots of cooling and power - Result: Data center

102 Google Data Center in Oregon

103 Google Data Center in Oregon Data centers (size of a football field)

104 Google Data Center in Oregon Data centers (size of a football field) 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)

105 Google Data Center Locations

106 Google Data Centers in the USA

107 Google Data Centers in Europe

108 Google Data Centers World Wide

109 Open Challenges

110 Open Challenges Can you manage thousands of racks effectively? - Cloud monitor systems (e.g., PlanetSeer [OSDI 04]) - Can you design more scalable data center network?

111 Open Challenges Can you manage thousands of racks effectively? - Cloud monitor systems (e.g., PlanetSeer [OSDI 04]) - Can you design more scalable data center network?

112 Open Challenges Can you manage thousands of racks effectively? - Cloud monitor systems (e.g., PlanetSeer [OSDI 04]) Can you make data center more scalable? - Scalable data center architecture (e.g., VL2 [SIGCOMM 09]) - Can you design more scalable data center network?

113 Open Challenges Can you manage thousands of racks effectively? - Cloud monitor systems (e.g., PlanetSeer [OSDI 04]) Can you make data center more scalable? - Scalable data center architecture (e.g., VL2 [SIGCOMM 09]) - Can you design more scalable data center network?

114 What Kinds of Challenges? Scalability Availability and reliability Security and privacy

115 What Kinds of Challenges? Scalability Availability and reliability Security and privacy

116 Availability & Reliability Is the cloud always there when you need it? - Service outages - Connectivity outages

117 Recent Cloud Disasters

118 Recent Cloud Disasters

119 Recent Cloud Disasters

120 Top10 Cloud Service Outages

121 Open Challenges Can you build a system to find out the root-cause when a service becomes unavailable? - Diagnosis systems (e.g., Sherlock [SIGCOMM 07])

122 Open Challenges Can you build a system to find out the root-cause when a service becomes unavailable? - Diagnosis systems (e.g., Sherlock [SIGCOMM 07]) - Accountable cloud (e.g., AVM [OSDI 10])

123 Open Challenges Can you build a system to find out the root-cause when a service becomes unavailable? - Diagnosis systems (e.g., Sherlock [SIGCOMM 07]) - Accountable cloud (e.g., AVM [OSDI 10])

124 Open Challenges Can you build a system to find out the root-cause when a service becomes unavailable? - Diagnosis systems (e.g., Sherlock [SIGCOMM 07]) - Accountable cloud (e.g., AVM [OSDI 10]) Can you propose an approach to make the clouds more robust? - Fault tolerate systems (e.g., F10 [NSDI 13])

125 Open Challenges Can you build a system to find out the root-cause when a service becomes unavailable? - Diagnosis systems (e.g., Sherlock [SIGCOMM 07]) - Accountable cloud (e.g., AVM [OSDI 10]) Can you propose an approach to make the clouds more robust? - Fault tolerate systems (e.g., F10 [NSDI 13])

126 What Kinds of Challenges? Scalability Availability and reliability Security and privacy

127 What Kinds of Challenges? Scalability Availability and reliability Security and privacy

128 Security & Privacy Compromised your cloud accounts - Hacker does not need to break into your home to steal all your private data, if he can break or guess your cloud password - Even worse, if hack who cracks your Facebook account get into your accounts everywhere online

129 Security & Privacy Compromised your cloud accounts - Hacker does not need to break into your home to steal all your private data, if he can break or guess your cloud password - Even worse, if hack who cracks your Facebook account get into your accounts everywhere online You do not know if the cloud providers read your private data

130 Open Challenges Can you build a system to preserve the privacy of your data on the clouds? - MAC for MapReduce (e.g., Airavat [NSDI 10]) - Trusted storage (e.g., Depot [OSDI 10])

131 Open Challenges Can you build a system to preserve the privacy of your data on the clouds? - MAC for MapReduce (e.g., Airavat [NSDI 10]) - Trusted storage (e.g., Depot [OSDI 10])

132 Open Challenges Can you build a system to preserve the privacy of your data on the clouds? - MAC for MapReduce (e.g., Airavat [NSDI 10]) - Trusted storage (e.g., Depot [OSDI 10])

133 Open Challenges Can you build a system to preserve the privacy of your data on the clouds? - MAC for MapReduce (e.g., Airavat [NSDI 10]) - Trusted storage (e.g., Depot [OSDI 10]) Can you propose an approach to verify if the cloud provider modifies your data? - Trusted cloud computing (e.g., Excalibur [USENIX Sec 12])

134 Open Challenges Can you build a system to preserve the privacy of your data on the clouds? - MAC for MapReduce (e.g., Airavat [NSDI 10]) - Trusted storage (e.g., Depot [OSDI 10]) Can you propose an approach to verify if the cloud provider modifies your data? - Trusted cloud computing (e.g., Excalibur [USENIX Sec 12])

135 More Risks?

136 More Risks? EverClouds is a project of DeDiS group (Bryan is PI): - aims to solve tricky cloud security problems (e.g., timing channels) - tries to make the clouds more reliable (e.g., failure detection) We already have some efforts: - SRA: A cloud structural reliability auditing system (submitted) - irec: A cloud independence recommender system (HotDep 13) - P-SRA: A privacy-preserving structural-reliability auditor (CCSW 13) - Timing channel control with provider-enforced deterministic execution (CCSW 10)

137 Lecture Outline Cloud Computing Challenges in the Clouds A Concrete Cloud Reliability Case

138 Lecture Outline Cloud Computing Challenges in the Clouds A Concrete Cloud Reliability Case

139 Realistic Problem

140 Realistic Problem Correlated failures resulting from EBS due to bugs in one EBS server Summary of the October 22, 2012 AWS Service Event in the US-East Region We d like to share more about the service event that occurred on Monday, October 22nd in the US-East Region. We have now completed the analysis of the events that affected AWS customers, and we want to describe what happened, our understanding of how customers were affected, and what we are doing to prevent a similar issue from occurring in the future. The Primary Event and the Impact to Amazon Elastic Block Store (EBS) and Amazon Elastic Compute Cloud (EC2)

141 Realistic Problem Elastic Compute Cloud (EC2) Elastic Block Store (EBS)

142 Realistic Problem Elastic Block Store (EBS)

143 Realistic Problem VM1 VM2 VM1 VM3 VM1 VM2 VM3 VM Elastic Block Store (EBS)

144 Realistic Problem VM1 VM2 VM1 VM3 VM1 VM2 VM3 VM EBS Server1 EBS Server2

145 Realistic Problem VM1 VM2 VM1 VM3 VM1 VM2 VM3 VM EBS Server1 EBS Server2

146 Realistic Problem VM1 VM2 VM1 VM3 VM1 VM2 VM3 VM EBS Server1 EBS Server2

147 Realistic Problem VM1 VM2 VM1 VM3 VM1 VM2 VM3 VM EBS Server1 EBS Server2

148 System We Need #1: Dependency collections #2: Dependency representation #3: Efficient auditing

149 System We Need #1: Dependency collections #2: Dependency representation #3: Efficient auditing

150 Dependency Data Collections Reuse existing data collection tools: - Convert the outputs to uniform format. - Three types of format: NET, HW and SW. Our defined format Type Network Hardware Software Dependency Expression <src= S dst= D route= x,y,z /> <hw= H type= T dep= x /> <pgm= S hw= H dep= x,y,z />

151 Dependency Data Collections

152 Dependency Data Collections NSDMiner DepDB

153 Dependency Data Collections NSDMiner DepDB

154 Dependency Data Collections NSDMiner DepDB

155 <src= S1 dst= Internet route= ToR1,Core1 /> <src= S1 dst= Internet route= ToR1,Core2 /> <src= S2 dst= Internet route= ToR1,Core1 /> <src= S2 dst= Internet route= ToR1,Core2 /> NSDMiner DepDB

156 <src= S1 dst= Internet route= ToR1,Core1 /> <src= S1 dst= Internet route= ToR1,Core2 /> <src= S2 dst= Internet route= ToR1,Core1 /> <src= S2 dst= Internet route= ToR1,Core2 /> NSDMiner DepDB

157 System We Need #1: Dependency collections #2: Dependency representation #3: Efficient auditing

158 Example Redundancy

159 Example Redundancy

160 Example Redundancy SW HW NET

161 Building Fault Graph Top-to-Bottom SW HW NET

162 Step1: Root Node Redundancy configuration fails

163 Step2: Server Nodes Redundancy configuration fails Server 1 fails Server 2 fails

164 Step2: Server Nodes Redundancy configuration fails Server 1 fails Server 2 fails AND gate: all the sublayer nodes fail, the upper layer node fails

165 Step3: Dependency Nodes Redundancy configuration fails Server 1 fails Server 2 fails +" +" HW fails Net fails SW fails Net fails SW fails HW fails

166 Step3: Dependency Nodes Redundancy configuration fails Server 1 fails Server 2 fails +" +" HW fails Net fails SW fails Net fails SW fails HW fails OR gate: one of the sublayer nodes fails, the upper layer node fails

167 Step4: Hardware Dependency Redundancy configuration fails Server 1 fails Server 2 fails +" +" HW fails Net fails SW fails Net fails SW fails HW fails +" +" CPU1 Disk1 CPU2 Disk2

168 Step5: Network Dependency Redundancy configuration fails Server 1 fails Server 2 fails +" +" HW fails Net fails SW fails Net fails SW fails HW fails +" +" CPU1 Disk1 Path1 Path2 CPU2 Disk2 Core1 +" +" ToR1 Core2

169 Step6: Software Dependency Redundancy configuration fails Server 1 fails Server 2 fails +" +" HW fails Net fails SW fails Net fails SW fails HW fails +" CPU1 Disk1 Path1 Path2 Riak +" Query CPU2 +" Disk2 +" +" +" +" Core1 ToR1 Core2 libsvnl libc6 libccl

170 System We Need #1: Dependency collections #2: Dependency representation #3: Efficient auditing

171 Efficient Auditing Two algorithms balancing cost and accuracy: - Minimal fault set algorithm - Failure sampling algorithm

172 Efficient Auditing Two algorithms balancing cost and accuracy: - Minimal fault set algorithm - Failure sampling algorithm

173 Minimal Fault Set Algorithm Traditional algorithm in safety engineering - Exponential complexity (NP-hard) We are the first to apply it in Cloud area: - Analyzing a fat tree with 30,528 with ~40 hours

174 Minimal Fault Set Algorithm Traditional algorithm in safety engineering - Exponential complexity (NP-hard) We are the first to apply it in Cloud area: - Analyzing a fat tree with 30,528 with ~40 hours We propose efficient failure sampling algorithm.

175 Failure Sampling Algorithm Redundancy configuration fails Server 1 fails Server 2 fails +" +" Server1 s HW fails Switch1 fails Server2 s HW fails

176 Failure Sampling Algorithm Redundancy configuration fails Server 1 fails Server 2 fails +" +" Server1 s HW fails Switch1 fails Server2 s HW fails 1 or 0 1 or 0 1 or 0

177 Failure Sampling Algorithm Redundancy configuration fails 1 or 0? Server 1 fails Server 2 fails +" +" Server1 s HW fails Switch1 fails Server2 s HW fails 1 or 0 1 or 0 1 or 0

178 Failure Sampling Algorithm Redundancy configuration fails 1 or 0? Server 1 fails Server 2 fails +" +" Server1 s HW fails Switch1 fails Server2 s HW fails 1 or 0 1 or 0 1 or 0 Fault Sets

179 The 1st Sampling Round Redundancy configuration fails Server 1 fails Server 2 fails +" +" Server1 s HW fails Switch1 fails Server2 s HW fails Fault Sets

180 The 1st Sampling Round Redundancy configuration fails Server 1 fails Server 2 fails +" +" Server1 s HW fails Switch1 fails Server2 s HW fails 1 or 0 1 or 0 1 or 0 Fault Sets

181 The 1st Sampling Round Redundancy configuration fails Server 1 fails Server 2 fails +" +" Server1 s HW fails Switch1 fails Fault Sets Server2 s HW fails

182 The 1st Sampling Round Redundancy configuration fails Server 1 fails Server 2 fails +" +" Server1 s HW fails Switch1 fails Fault Sets Server2 s HW fails

183 The 1st Sampling Round Redundancy configuration fails Server 1 fails Server 2 fails +" +" Server1 s HW fails Switch1 fails Fault Sets Server2 s HW fails {Server1 s HW, Server2 s HW}

184 The 2nd Sampling Round Redundancy configuration fails Server 1 fails Server 2 fails +" +" Server1 s HW fails Switch1 fails Server2 s HW fails Fault Sets {Server1 s HW, Server2 s HW}

185 The 2nd Sampling Round Redundancy configuration fails Server 1 fails Server 2 fails +" +" Server1 s HW fails Switch1 fails Server2 s HW fails 1 or 0 1 or 0 1 or 0 Fault Sets {Server1 s HW, Server2 s HW}

186 The 2nd Sampling Round Redundancy configuration fails Server 1 fails Server 2 fails +" +" Server1 s HW fails Switch1 fails Fault Sets Server2 s HW fails {Server1 s HW, Server2 s HW}

187 The 2nd Sampling Round Redundancy configuration fails Server 1 fails Server 2 fails +" +" Server1 s HW fails Switch1 fails Fault Sets Server2 s HW fails {Server1 s HW, Server2 s HW}

188 The 3rd Sampling Round Redundancy configuration fails Server 1 fails Server 2 fails +" +" Server1 s HW fails Switch1 fails Server2 s HW fails Fault Sets {Server1 s HW, Server2 s HW}

189 The 3rd Sampling Round Redundancy configuration fails Server 1 fails Server 2 fails +" +" Server1 s HW fails Switch1 fails Fault Sets Server2 s HW fails {Server1 s HW, Server2 s HW}

190 The 3rd Sampling Round Redundancy configuration fails Server 1 fails Server 2 fails +" +" Server1 s HW fails Switch1 fails Fault Sets Server2 s HW fails {Server1 s HW, Server2 s HW}

191 The 3rd Sampling Round Redundancy configuration fails Server 1 fails Server 2 fails +" +" Server1 s HW fails Switch1 fails Fault Sets Server2 s HW fails {Server1 s HW, Server2 s HW} {Switch1}

192 After Many (e.g., 10 7 ) Rounds Fault Sets {Server1 s HW, Server2 s HW} {Switch1} {Switch1, Server2 s HW} {Switch1} {Switch1, Server2 s HW}......

193 Size-Based Ranking Fault Sets {Server1 s HW, Server2 s HW} {Switch1} {Switch1, Server2 s HW} {Switch1} {Switch1, Server2 s HW}......

194 Size-Based Ranking Fault Sets {Switch1} {Switch1} {Switch1, Server2 s HW} {Switch1, Server2 s HW} {Server1 s HW, Server2 s HW}......

195 Thanks! Questions?