Topic 2: Cloud Computing Paradigms
1 like1,767 views
This document discusses cloud computing paradigms. It outlines the history and growth of cloud computing, driven by large internet companies developing scalable infrastructure and data centers. It describes advantages for both service providers and end users, such as simplified installation and maintenance, centralized updates, and on-demand scaling. New applications like mobile and analytics are discussed. Cloud providers operate under a utility computing model of low commitment and pay-as-you-go services.
Topic 2: Cloud Computing Paradigms
- 1. 2: Cloud Computing Paradigms Zubair Nabi zubair.nabi@itu.edu.pk April 17, 2013 Zubair Nabi 2: Cloud Computing Paradigms April 17, 2013 1 / 22
- 2. Outline 1 Introduction 2 Cloud service providers 3 Utility Computing 4 Economics 5 Challenges Zubair Nabi 2: Cloud Computing Paradigms April 17, 2013 2 / 22
- 3. Outline 1 Introduction 2 Cloud service providers 3 Utility Computing 4 Economics 5 Challenges Zubair Nabi 2: Cloud Computing Paradigms April 17, 2013 3 / 22
- 4. Cloud computing A realization of utility computing in which computation, storage, and services are offered as a metered service Zubair Nabi 2: Cloud Computing Paradigms April 17, 2013 4 / 22
- 5. Cloud computing A realization of utility computing in which computation, storage, and services are offered as a metered service Encompasses applications delivered as services over the Internet and hardware and software in the datacenters that enable those services Zubair Nabi 2: Cloud Computing Paradigms April 17, 2013 4 / 22
- 6. Cloud computing A realization of utility computing in which computation, storage, and services are offered as a metered service Encompasses applications delivered as services over the Internet and hardware and software in the datacenters that enable those services Software as a Service (SaaS) Zubair Nabi 2: Cloud Computing Paradigms April 17, 2013 4 / 22
- 7. Cloud computing A realization of utility computing in which computation, storage, and services are offered as a metered service Encompasses applications delivered as services over the Internet and hardware and software in the datacenters that enable those services Software as a Service (SaaS) Public Cloud: If available to the public as a pay-as-you-go model, e.g. Amazon Web Services, Google AppEngine, and Microsoft Azure Zubair Nabi 2: Cloud Computing Paradigms April 17, 2013 4 / 22
- 8. Cloud computing A realization of utility computing in which computation, storage, and services are offered as a metered service Encompasses applications delivered as services over the Internet and hardware and software in the datacenters that enable those services Software as a Service (SaaS) Public Cloud: If available to the public as a pay-as-you-go model, e.g. Amazon Web Services, Google AppEngine, and Microsoft Azure Private Cloud: Internal datacenters of an organization that are not publicly accessible Zubair Nabi 2: Cloud Computing Paradigms April 17, 2013 4 / 22
- 9. Advantages Advantages to both service providers and end users Zubair Nabi 2: Cloud Computing Paradigms April 17, 2013 5 / 22
- 10. Advantages Advantages to both service providers and end users 1 Service providers: Simplified software installation and maintenance Zubair Nabi 2: Cloud Computing Paradigms April 17, 2013 5 / 22
- 11. Advantages Advantages to both service providers and end users 1 Service providers: Simplified software installation and maintenance Centralized control over versioning Zubair Nabi 2: Cloud Computing Paradigms April 17, 2013 5 / 22
- 12. Advantages Advantages to both service providers and end users 1 Service providers: Simplified software installation and maintenance Centralized control over versioning No need to build, provision, and maintain a datacenter Zubair Nabi 2: Cloud Computing Paradigms April 17, 2013 5 / 22
- 13. Advantages Advantages to both service providers and end users 1 Service providers: Simplified software installation and maintenance Centralized control over versioning No need to build, provision, and maintain a datacenter On the fly scaling Zubair Nabi 2: Cloud Computing Paradigms April 17, 2013 5 / 22
- 14. Advantages Advantages to both service providers and end users 1 Service providers: Simplified software installation and maintenance Centralized control over versioning No need to build, provision, and maintain a datacenter On the fly scaling 2 End users: “Anytime, anywhere” access Zubair Nabi 2: Cloud Computing Paradigms April 17, 2013 5 / 22
- 15. Advantages Advantages to both service providers and end users 1 Service providers: Simplified software installation and maintenance Centralized control over versioning No need to build, provision, and maintain a datacenter On the fly scaling 2 End users: “Anytime, anywhere” access Share data and collaborate easily Zubair Nabi 2: Cloud Computing Paradigms April 17, 2013 5 / 22
- 16. Advantages Advantages to both service providers and end users 1 Service providers: Simplified software installation and maintenance Centralized control over versioning No need to build, provision, and maintain a datacenter On the fly scaling 2 End users: “Anytime, anywhere” access Share data and collaborate easily Safeguard data stored in the infrastructure (debatable) Zubair Nabi 2: Cloud Computing Paradigms April 17, 2013 5 / 22
- 17. Outline 1 Introduction 2 Cloud service providers 3 Utility Computing 4 Economics 5 Challenges Zubair Nabi 2: Cloud Computing Paradigms April 17, 2013 6 / 22
- 18. History Phenomenal growth of Web services in late 90s and early 2000s Zubair Nabi 2: Cloud Computing Paradigms April 17, 2013 7 / 22
- 19. History Phenomenal growth of Web services in late 90s and early 2000s Large Internet companies, including Amazon, eBay, Google, Microsoft, Yahoo, etc., already had massive infrastructure Zubair Nabi 2: Cloud Computing Paradigms April 17, 2013 7 / 22
- 20. History Phenomenal growth of Web services in late 90s and early 2000s Large Internet companies, including Amazon, eBay, Google, Microsoft, Yahoo, etc., already had massive infrastructure To keep up with demand, these companies also developed scalable software infrastructure (think MapReduce, GFS, BigTable, Dynamo, etc.) Zubair Nabi 2: Cloud Computing Paradigms April 17, 2013 7 / 22
- 21. History Phenomenal growth of Web services in late 90s and early 2000s Large Internet companies, including Amazon, eBay, Google, Microsoft, Yahoo, etc., already had massive infrastructure To keep up with demand, these companies also developed scalable software infrastructure (think MapReduce, GFS, BigTable, Dynamo, etc.) They also acquired the operational expertise to deter potential physical and electronic attacks Zubair Nabi 2: Cloud Computing Paradigms April 17, 2013 7 / 22
- 22. History Phenomenal growth of Web services in late 90s and early 2000s Large Internet companies, including Amazon, eBay, Google, Microsoft, Yahoo, etc., already had massive infrastructure To keep up with demand, these companies also developed scalable software infrastructure (think MapReduce, GFS, BigTable, Dynamo, etc.) They also acquired the operational expertise to deter potential physical and electronic attacks Therefore, they had already created extremely large datacenters to leverage statistical multiplexing and bulk purchasing of infrastructure Zubair Nabi 2: Cloud Computing Paradigms April 17, 2013 7 / 22
- 23. Incentive for providers Incentives include revenue, leveraging existing investment, defending a franchise, attacking an incumbent, leveraging customer relationships, and becoming a platform Zubair Nabi 2: Cloud Computing Paradigms April 17, 2013 8 / 22
- 24. Incentive for providers Incentives include revenue, leveraging existing investment, defending a franchise, attacking an incumbent, leveraging customer relationships, and becoming a platform Data centers are being established in seemingly arbitrary locations Zubair Nabi 2: Cloud Computing Paradigms April 17, 2013 8 / 22
- 25. Incentive for providers Incentives include revenue, leveraging existing investment, defending a franchise, attacking an incumbent, leveraging customer relationships, and becoming a platform Data centers are being established in seemingly arbitrary locations Reasons for choosing a location include costs of electricity, cooling, labour, property, and taxes Zubair Nabi 2: Cloud Computing Paradigms April 17, 2013 8 / 22
- 26. Incentive for providers Incentives include revenue, leveraging existing investment, defending a franchise, attacking an incumbent, leveraging customer relationships, and becoming a platform Data centers are being established in seemingly arbitrary locations Reasons for choosing a location include costs of electricity, cooling, labour, property, and taxes Cooling and electricity account for 1/3rd of all costs! Zubair Nabi 2: Cloud Computing Paradigms April 17, 2013 8 / 22
- 27. Incentive for providers Incentives include revenue, leveraging existing investment, defending a franchise, attacking an incumbent, leveraging customer relationships, and becoming a platform Data centers are being established in seemingly arbitrary locations Reasons for choosing a location include costs of electricity, cooling, labour, property, and taxes Cooling and electricity account for 1/3rd of all costs! Cheaper to ship data over fiber optic cables than to ship electricity over high-voltage transmission lines Zubair Nabi 2: Cloud Computing Paradigms April 17, 2013 8 / 22
- 28. New technology trends and business models “High-touch, high-margin, high-commitment” provisioning of service to “low-touch, low-margin, low-commitment” Zubair Nabi 2: Cloud Computing Paradigms April 17, 2013 9 / 22
- 29. New technology trends and business models “High-touch, high-margin, high-commitment” provisioning of service to “low-touch, low-margin, low-commitment” For instance: Payment model in Web 1.0: Contractual arrangement with a payment processing service such as VeriSign or Authorize.net; making it hard for small businesses to accept credit card payment online Zubair Nabi 2: Cloud Computing Paradigms April 17, 2013 9 / 22
- 30. New technology trends and business models “High-touch, high-margin, high-commitment” provisioning of service to “low-touch, low-margin, low-commitment” For instance: Payment model in Web 1.0: Contractual arrangement with a payment processing service such as VeriSign or Authorize.net; making it hard for small businesses to accept credit card payment online Web 2.0: With PayPal-like services anyone can sign up and accept credit payments without a contract and a long-term commitment Zubair Nabi 2: Cloud Computing Paradigms April 17, 2013 9 / 22
- 31. New technology trends and business models “High-touch, high-margin, high-commitment” provisioning of service to “low-touch, low-margin, low-commitment” For instance: Payment model in Web 1.0: Contractual arrangement with a payment processing service such as VeriSign or Authorize.net; making it hard for small businesses to accept credit card payment online Web 2.0: With PayPal-like services anyone can sign up and accept credit payments without a contract and a long-term commitment Another example: Ad revenue model in Web 1.0: Set up a relationship with an ad placement company, such as DoubleClick Zubair Nabi 2: Cloud Computing Paradigms April 17, 2013 9 / 22
- 32. New technology trends and business models “High-touch, high-margin, high-commitment” provisioning of service to “low-touch, low-margin, low-commitment” For instance: Payment model in Web 1.0: Contractual arrangement with a payment processing service such as VeriSign or Authorize.net; making it hard for small businesses to accept credit card payment online Web 2.0: With PayPal-like services anyone can sign up and accept credit payments without a contract and a long-term commitment Another example: Ad revenue model in Web 1.0: Set up a relationship with an ad placement company, such as DoubleClick Web 2.0: Use Google AdSense Zubair Nabi 2: Cloud Computing Paradigms April 17, 2013 9 / 22
- 33. New technology trends and business models “High-touch, high-margin, high-commitment” provisioning of service to “low-touch, low-margin, low-commitment” For instance: Payment model in Web 1.0: Contractual arrangement with a payment processing service such as VeriSign or Authorize.net; making it hard for small businesses to accept credit card payment online Web 2.0: With PayPal-like services anyone can sign up and accept credit payments without a contract and a long-term commitment Another example: Ad revenue model in Web 1.0: Set up a relationship with an ad placement company, such as DoubleClick Web 2.0: Use Google AdSense This same model was used by Amazon Web Services in 2006: pay-as-you-go computing with no contract, with the only requirement being a credit card Zubair Nabi 2: Cloud Computing Paradigms April 17, 2013 9 / 22
- 34. New applications Mobile applications: Require high availability and rely on large data sets that are most conveniently hosted in large datacenters Zubair Nabi 2: Cloud Computing Paradigms April 17, 2013 10 / 22
- 35. New applications Mobile applications: Require high availability and rely on large data sets that are most conveniently hosted in large datacenters Parallel batch processing: Analytics jobs that analyze terabytes of data and can take hours to finish can leverage the “cost associativity” of the cloud Zubair Nabi 2: Cloud Computing Paradigms April 17, 2013 10 / 22
- 36. New applications Mobile applications: Require high availability and rely on large data sets that are most conveniently hosted in large datacenters Parallel batch processing: Analytics jobs that analyze terabytes of data and can take hours to finish can leverage the “cost associativity” of the cloud Business analytics: Understanding customers, supply chains, buying habits, ranking, and so on Zubair Nabi 2: Cloud Computing Paradigms April 17, 2013 10 / 22
- 37. New applications Mobile applications: Require high availability and rely on large data sets that are most conveniently hosted in large datacenters Parallel batch processing: Analytics jobs that analyze terabytes of data and can take hours to finish can leverage the “cost associativity” of the cloud Business analytics: Understanding customers, supply chains, buying habits, ranking, and so on Computation offloading: Compute-intensive tasks are offloaded to the cloud. For instance, Matlab, Mathematica, image rendering, 3D animations, etc. Zubair Nabi 2: Cloud Computing Paradigms April 17, 2013 10 / 22
- 38. Outline 1 Introduction 2 Cloud service providers 3 Utility Computing 4 Economics 5 Challenges Zubair Nabi 2: Cloud Computing Paradigms April 17, 2013 11 / 22
- 39. Classes of utility computing Every application needs computation, storage, and quite possibly communication Zubair Nabi 2: Cloud Computing Paradigms April 17, 2013 12 / 22
- 40. Classes of utility computing Every application needs computation, storage, and quite possibly communication These resources need to be virtualized to achieve elasticity and the illusion of infinite capacity Zubair Nabi 2: Cloud Computing Paradigms April 17, 2013 12 / 22
- 41. Classes of utility computing Every application needs computation, storage, and quite possibly communication These resources need to be virtualized to achieve elasticity and the illusion of infinite capacity The details of statistical multiplexing and sharing is abstracted away from the programmer Zubair Nabi 2: Cloud Computing Paradigms April 17, 2013 12 / 22
- 42. Classes of utility computing Every application needs computation, storage, and quite possibly communication These resources need to be virtualized to achieve elasticity and the illusion of infinite capacity The details of statistical multiplexing and sharing is abstracted away from the programmer Different utility computing offerings can be distinguished on the basis of the abstraction presented to the programmer Zubair Nabi 2: Cloud Computing Paradigms April 17, 2013 12 / 22
- 43. Bare metal hardware abstraction An instance looks like physical hardware Zubair Nabi 2: Cloud Computing Paradigms April 17, 2013 13 / 22
- 44. Bare metal hardware abstraction An instance looks like physical hardware Programmers control the entire software stack from the kernel upwards Zubair Nabi 2: Cloud Computing Paradigms April 17, 2013 13 / 22
- 45. Bare metal hardware abstraction An instance looks like physical hardware Programmers control the entire software stack from the kernel upwards Employed by Amazon EC2 Zubair Nabi 2: Cloud Computing Paradigms April 17, 2013 13 / 22
- 46. Bare metal hardware abstraction An instance looks like physical hardware Programmers control the entire software stack from the kernel upwards Employed by Amazon EC2 A very thin API is exposed to request and configure virtualized hardware Zubair Nabi 2: Cloud Computing Paradigms April 17, 2013 13 / 22
- 47. Bare metal hardware abstraction An instance looks like physical hardware Programmers control the entire software stack from the kernel upwards Employed by Amazon EC2 A very thin API is exposed to request and configure virtualized hardware No bar on the kinds of applications that can be hosted Zubair Nabi 2: Cloud Computing Paradigms April 17, 2013 13 / 22
- 48. Bare metal hardware abstraction An instance looks like physical hardware Programmers control the entire software stack from the kernel upwards Employed by Amazon EC2 A very thin API is exposed to request and configure virtualized hardware No bar on the kinds of applications that can be hosted Low level virtualization, block-device storage, and IP-level connectivity allow developers to design any application Zubair Nabi 2: Cloud Computing Paradigms April 17, 2013 13 / 22
- 49. Bare metal hardware abstraction An instance looks like physical hardware Programmers control the entire software stack from the kernel upwards Employed by Amazon EC2 A very thin API is exposed to request and configure virtualized hardware No bar on the kinds of applications that can be hosted Low level virtualization, block-device storage, and IP-level connectivity allow developers to design any application On the downside, scalability and failover are application-dependent Zubair Nabi 2: Cloud Computing Paradigms April 17, 2013 13 / 22
- 50. Domain-specific platform Target traditional web applications Zubair Nabi 2: Cloud Computing Paradigms April 17, 2013 14 / 22
- 51. Domain-specific platform Target traditional web applications Enforce an application structure of clean separation between a stateless computation tier and a stateful storage tier Zubair Nabi 2: Cloud Computing Paradigms April 17, 2013 14 / 22
- 52. Domain-specific platform Target traditional web applications Enforce an application structure of clean separation between a stateless computation tier and a stateful storage tier Employed by Google AppEngine Zubair Nabi 2: Cloud Computing Paradigms April 17, 2013 14 / 22
- 53. Domain-specific platform Target traditional web applications Enforce an application structure of clean separation between a stateless computation tier and a stateful storage tier Employed by Google AppEngine Applications are expected to be request-reply based Zubair Nabi 2: Cloud Computing Paradigms April 17, 2013 14 / 22
- 54. Domain-specific platform Target traditional web applications Enforce an application structure of clean separation between a stateless computation tier and a stateful storage tier Employed by Google AppEngine Applications are expected to be request-reply based In contrast to the bare metal hardware abstraction, enable automatic scaling and high-availability mechanisms Zubair Nabi 2: Cloud Computing Paradigms April 17, 2013 14 / 22
- 55. Domain-specific platform Target traditional web applications Enforce an application structure of clean separation between a stateless computation tier and a stateful storage tier Employed by Google AppEngine Applications are expected to be request-reply based In contrast to the bare metal hardware abstraction, enable automatic scaling and high-availability mechanisms Not suitable for general-purpose computing Zubair Nabi 2: Cloud Computing Paradigms April 17, 2013 14 / 22
- 56. Hybrid Offer a sweet spot between flexibility and programmer convenience Zubair Nabi 2: Cloud Computing Paradigms April 17, 2013 15 / 22
- 57. Hybrid Offer a sweet spot between flexibility and programmer convenience Offered by Microsoft’s Azure Zubair Nabi 2: Cloud Computing Paradigms April 17, 2013 15 / 22
- 58. Hybrid Offer a sweet spot between flexibility and programmer convenience Offered by Microsoft’s Azure Applications are written using .NET libraries and compiled to the Common Language Runtime (A language-independent management environment) Zubair Nabi 2: Cloud Computing Paradigms April 17, 2013 15 / 22
- 59. Hybrid Offer a sweet spot between flexibility and programmer convenience Offered by Microsoft’s Azure Applications are written using .NET libraries and compiled to the Common Language Runtime (A language-independent management environment) Supports general purpose computing Zubair Nabi 2: Cloud Computing Paradigms April 17, 2013 15 / 22
- 60. Hybrid Offer a sweet spot between flexibility and programmer convenience Offered by Microsoft’s Azure Applications are written using .NET libraries and compiled to the Common Language Runtime (A language-independent management environment) Supports general purpose computing Users have control over the choice of language but not the underlying OS or runtime Zubair Nabi 2: Cloud Computing Paradigms April 17, 2013 15 / 22
- 61. Hybrid Offer a sweet spot between flexibility and programmer convenience Offered by Microsoft’s Azure Applications are written using .NET libraries and compiled to the Common Language Runtime (A language-independent management environment) Supports general purpose computing Users have control over the choice of language but not the underlying OS or runtime Provide some degree of automatic failover and scalability but require some help from the developer in the form of declaration of some application properties Zubair Nabi 2: Cloud Computing Paradigms April 17, 2013 15 / 22
- 62. Outline 1 Introduction 2 Cloud service providers 3 Utility Computing 4 Economics 5 Challenges Zubair Nabi 2: Cloud Computing Paradigms April 17, 2013 16 / 22
- 63. Elasticity Pay-as-you-go model: Only pay for what you use Zubair Nabi 2: Cloud Computing Paradigms April 17, 2013 17 / 22
- 64. Elasticity Pay-as-you-go model: Only pay for what you use Add or remove resources at a fine grain (such as one server at a time) with minimal lead time Zubair Nabi 2: Cloud Computing Paradigms April 17, 2013 17 / 22
- 65. Elasticity Pay-as-you-go model: Only pay for what you use Add or remove resources at a fine grain (such as one server at a time) with minimal lead time Useful for traffic spikes such as “Black Friday” Zubair Nabi 2: Cloud Computing Paradigms April 17, 2013 17 / 22
- 66. Elasticity Pay-as-you-go model: Only pay for what you use Add or remove resources at a fine grain (such as one server at a time) with minimal lead time Useful for traffic spikes such as “Black Friday” Over time, hardware costs come down and vendors acquire updated hardware. Thus, benefiting the tenant Zubair Nabi 2: Cloud Computing Paradigms April 17, 2013 17 / 22
- 67. Reasons for companies to migrate to the cloud Pay separately per resource: Pay proportional to resource requirements Zubair Nabi 2: Cloud Computing Paradigms April 17, 2013 18 / 22
- 68. Reasons for companies to migrate to the cloud Pay separately per resource: Pay proportional to resource requirements Power, cooling, and physical plant costs: Cost of electricity and cooling already factored in Zubair Nabi 2: Cloud Computing Paradigms April 17, 2013 18 / 22
- 69. Reasons for companies to migrate to the cloud Pay separately per resource: Pay proportional to resource requirements Power, cooling, and physical plant costs: Cost of electricity and cooling already factored in Man-power costs: No need to employ sysadmins Zubair Nabi 2: Cloud Computing Paradigms April 17, 2013 18 / 22
- 70. Reasons for companies to migrate to the cloud Pay separately per resource: Pay proportional to resource requirements Power, cooling, and physical plant costs: Cost of electricity and cooling already factored in Man-power costs: No need to employ sysadmins Operational costs: Low-level upgrades and software patches responsibility of the provider Zubair Nabi 2: Cloud Computing Paradigms April 17, 2013 18 / 22
- 71. Outline 1 Introduction 2 Cloud service providers 3 Utility Computing 4 Economics 5 Challenges Zubair Nabi 2: Cloud Computing Paradigms April 17, 2013 19 / 22
- 72. Obstacles 1 Service availability: Possibility of cloud outage Zubair Nabi 2: Cloud Computing Paradigms April 17, 2013 20 / 22
- 73. Obstacles 1 Service availability: Possibility of cloud outage 2 Data lock-in: Reliance on cloud specific APIs Zubair Nabi 2: Cloud Computing Paradigms April 17, 2013 20 / 22
- 74. Obstacles 1 Service availability: Possibility of cloud outage 2 Data lock-in: Reliance on cloud specific APIs 3 Security: Requires strong encrypted storage, VLANs, and network middleboxes (firewalls, etc.) Zubair Nabi 2: Cloud Computing Paradigms April 17, 2013 20 / 22
- 75. Obstacles 1 Service availability: Possibility of cloud outage 2 Data lock-in: Reliance on cloud specific APIs 3 Security: Requires strong encrypted storage, VLANs, and network middleboxes (firewalls, etc.) 4 Data transfer bottlenecks: Moving large amounts of data in and out is expensive Zubair Nabi 2: Cloud Computing Paradigms April 17, 2013 20 / 22
- 76. Obstacles 1 Service availability: Possibility of cloud outage 2 Data lock-in: Reliance on cloud specific APIs 3 Security: Requires strong encrypted storage, VLANs, and network middleboxes (firewalls, etc.) 4 Data transfer bottlenecks: Moving large amounts of data in and out is expensive 5 Performance unpredictability: Resource sharing between applications Zubair Nabi 2: Cloud Computing Paradigms April 17, 2013 20 / 22
- 77. Obstacles (2) 6 Scalable storage: No standard model to arbitrarily scale storage up and down on-demand while ensuring data durability and high availability Zubair Nabi 2: Cloud Computing Paradigms April 17, 2013 21 / 22
- 78. Obstacles (2) 6 Scalable storage: No standard model to arbitrarily scale storage up and down on-demand while ensuring data durability and high availability 7 Bugs in large-scale distributed systems: Hard to debug large-scale applications in full deployment Zubair Nabi 2: Cloud Computing Paradigms April 17, 2013 21 / 22
- 79. Obstacles (2) 6 Scalable storage: No standard model to arbitrarily scale storage up and down on-demand while ensuring data durability and high availability 7 Bugs in large-scale distributed systems: Hard to debug large-scale applications in full deployment 8 Scaling quickly: Automatically scaling while conserving resources and money is an open ended problem Zubair Nabi 2: Cloud Computing Paradigms April 17, 2013 21 / 22
- 80. Obstacles (2) 6 Scalable storage: No standard model to arbitrarily scale storage up and down on-demand while ensuring data durability and high availability 7 Bugs in large-scale distributed systems: Hard to debug large-scale applications in full deployment 8 Scaling quickly: Automatically scaling while conserving resources and money is an open ended problem 9 Reputation fate sharing: Bad behaviour by one tenant can reflect badly on the rest Zubair Nabi 2: Cloud Computing Paradigms April 17, 2013 21 / 22
- 81. Obstacles (2) 6 Scalable storage: No standard model to arbitrarily scale storage up and down on-demand while ensuring data durability and high availability 7 Bugs in large-scale distributed systems: Hard to debug large-scale applications in full deployment 8 Scaling quickly: Automatically scaling while conserving resources and money is an open ended problem 9 Reputation fate sharing: Bad behaviour by one tenant can reflect badly on the rest 10 Software licensing: Gap between pay-as-you-go model and software licensing Zubair Nabi 2: Cloud Computing Paradigms April 17, 2013 21 / 22
- 82. References 1 Above the Clouds: A Berkeley View of Cloud Computing" by Michael Armbrust, Armando Fox, Rean Griffith, Anthony D. Joseph, Randy Katz, Andy Konwinski, Gunho Lee, David Patterson, Ariel Rabkin, Ion Stoica, and Matei Zaharia. Technical Report EECS-2009-28, EECS Department, University of California, Berkeley. Zubair Nabi 2: Cloud Computing Paradigms April 17, 2013 22 / 22