GRID ECONOMICS

1. GRID ECONOMICS Lecturer: Ph.D. Pham Tran Vu Students: Lam Hoang Vu - 00708218 Nguyen Quang Minh - 00708201

2. Outline • Introduction. • What is Grid economics? • Grid economics vision. • Players in Grid marketplace. • Economic/business models. • Grid practices: • Nimrod-G. • Q & A.

3. Introduction • Current using of Grid in enterprises: • In large enterprises: • Consolidating IT resources. • Improving the workflow within an enterprise. • In small and medium enterprises (SMEs): • Almost not consider. • Miss out many helpful Grid’s properties.

4. Introduction • Grid can supply: • Availability of on-demand resources. • When consumer requests unavailable resources. • Supplier can buy the resources on the Grid. • Easily access to resources. • Low cost ownership. • Resources that you can’t afford for possession. • Pay-for-use pricing model. • Only pay for the usage of resources.

5. Introduction • Grid can help SMEs: • Opportunities. • Independency. • Compete with large enterprises.

6. What is Grid economics? • Definition: • Applying economic principles to Grid computing. • Grid infrastructure to support commercial, economic Grids. • Service Level Agreements (SLAs). • Pricing and Capacity Planning. • Business models for the Grid.

7. What is Grid business? • Definition: • Business activities. • Through commercialized IT services. • Based on Grid computing. • Grid business model: • Defines a framework for creating new value chains.

8. Some Grid-economics Projects • European Grid Projects • GridEcon, http://www.gridecon.eu • BEinGRID, http://www.beingrid.eu/ • Gridbus,http://www.gridbus.org/ • American Grid Project • TeraGrid , http://www.teragrid.org/ • GT4, http://www.globus.org/ • Asian Grid Project • K*Grid,  http://www.gridcenter.or.kr/ (South Korea) • NAREGI, http://www.naregi.org/index_e.html (Japan) • CNGrid http://www.cngrid.org/ (China) • Australian Grid Project • GridBus, http://www.gridbus.org/

9. Example • Amazon EC2: • Allows the user to create virtual machines. • VM: • Initiate, run and monitor applications. • Price (Feb 2007): • $0.10 per instance-hour. • $0.20 per GB of transferred data. • $0.15 per GB-Month of Amazon S3 Storage. • 1.7Ghz Xeon CPU. 160GB local disk. • 1.75GB RAM. 250Mb/s network bandwidth.

10. Grid economics vision • Assume: • Grid needs a part called “Open market”: • Allows participants to: • Buy services. • Sell enhanced services at the same time.

11. Grid economics vision (cont.) • 3 existing technologies: • Service-oriented computing. • Virtualization of resources. • Network computing. • 1 missing: Economic-enhanced services • Economic risks and transaction opportunities. • Enlarge the access of existing Grid business. • Grid middleware  Economic-aware  or Market-aware

12. Grid economics architecture • Three layers: • Consumer. • Economic-enhanced service provider. • The basic resources provider.

13. Players in the Grid Marketplace • 2 key players: • Grid Service Providers (GSPs): producers. • Grid Resource Brokers (GRBs): consumers.

14. Players in the Grid Marketplace • The GSPs: • Make their resources Grid enabled by running software systems along with Grid Trading Services (GTS). • Interaction between GSPs and GRBs: • Through a Grid Market Directory (GMD). • For deciding service access price. • Economic models. • Interaction protocols.

15. Players in the Grid Marketplace • GRBs: • May invite bids from a number of GSPs. • Select those that offer: • Lowest service costs. • Meet their deadline and budget requirements. • GSPs: • May invite bids in an auction. • Offer services to the highest bidder as long as its objectives are met.

16. Business Models Commodity Market Model Posted Price Model Bargaining Model Tendering/Contract-Net Model Auction Model Bid-based Proportional Resource Sharing Model

17. Business Models 1.Commodity Market Model G Interaction between GSPs and users in a commodity market rid.

18. Business Models • Pricing schemes in a Commodity Market Model can be based on: • Flat fee. • Usage Duration (Time). • Subscription. • Demand and Supply-based.

19. Business Models • Deployment: • The users compose their application using higher-level Grid programming languages. • The resource broker (working for the user) can carry out the following steps for executing applications: • The broker identifies service providers. • It identifies suitable resources and establishes their prices (by interacting with GMD and GTS). • It selects resources that meet its utility function and objectives (lower cost and meet deadline requirements). • It uses resource services for job processing and issues payments as agreed.

20. Business Models 2. Posted Price Model Posted price model and resource trading

21. Business Models • The posted price model: • Is similar to the commodity market model, except that it advertises special offers. • The activities that are: • Grid Service Providers (GSPs) post their special offers and associated conditions etc. in Grid Market Directory. • Broker looks at GMD to identify if any of these posted services are available and fits its requirements. • Broker enquires (GSP) for availability of posted services. • Other steps are similar to those pointed out in commodity market model.

22. Business Models 3. Bargaining Model Bargaining for lower access price in their bid for minimizing computational cost.

23. Business Models • In the bargaining model: • Both brokers and GSPs have their own objective functions. • They negotiate with each other as long as their objectives are met.

24. Tender/Contract-Net Model • Being one of the most widely used models for service negotiation in a distributed problem solving environment. • This model bases on the contracting mechanism • Used by business to govern the exchange of goods and services • Finding an appropriate service provider to work on a given task.

25. Tender/Contract-Net Model (cont’) Deliver two concepts • Manager - a user / resource broker asks for a task to be solved. • Contractor–resource might be able to solve a task.

26. Tender/Contract-Net Model (cont’) • The process (of tender/contract-net) according to manager’s perspective • Consumer (broker)announces its requirements and invites bids from GSPs. • Interested GSPs evaluate the announcement and respond by submitting their bids. • Broker evaluates and awards the contract to the most appropriate GSP(s). • The broker and GSP communicate privately and use the resource (R).

27. Tender/Contract-Net Model (cont’) • The process (of tender/contract-net) according to contractor’s perspective • Receive tender announcements/advertisements (say in GMD). • Evaluate service capability. • Respond with bid. • Deliver service if bid is accepted. • Report results and bill the broker/user as per the usage and agreed bid.

28. Tender/Contract-Net Model (cont’) • Advantage • The brokers have many choices • e.g. if the selected GSP is unable to deliver a satisfactory service, the brokers can seek services of other GSPs. • Disadvantage • A less capable GSP can be awarded to a task (if a more capable GSP is busy). • Sometimes, manager may not receive bids (for reasons such as all potential GSPs are busy with other tasks).

29. Auction model • This model is proper for one-to-many negotiation between a service provider and many consumers within a set duration. • The key players participate in auctions: • resource owners (GSPs) • buyers (consumers/brokers) • auctioneers (who set the rules of auction)

30. Auction model (cont’) Auctions using external auctioneers

31. Auction model (cont’) • The auction process contains the following steps • GSPs announce their services and invite bids. • Brokers offer their bids (and they can see what other consumers offer if they like - depending on open/closed). • Step (b) goes on until no one is willing to bid higher price or auctioneer stops if the minimum price line is not met. • GSP offers service to the one who wins. • Consumer uses the resource.

32. Bid-based Proportional Resource Sharing Model This model is quite popular in cooperative problem-solving environments

33. Bid-based Proportional Resource Sharing Model (cont’) • The users are allocated credits, which they can use for having access to shared resource. • The value of each credit depends on the resource demand and the value that other users place on the resource at the time of usage.

34. Grid Practice • Nimrod-G • Developed at MonashUniversity, Melbourne (Buyya, Abramson, Giddy) • GridEcon • A European Union funded project on Grid Economics and Business Models

35. Nimrod-G • Nimrod/G architecture for resource management and scheduling bases on a computational economic model. • It uses the Globus toolkit services and can be easily extended to operate with any other emerging grid middleware services. • Supports deadline and budget-constrained algorithms. • Supports GUI tools and declarative programming language.

36. Nimrod-G (cont’) • The Nimrod-G architecture contains the following key components: • Client or User Station • Parametric Engine • Scheduler • Dispatcher • Job-Wrapper

37. Nimrod-G (cont’) • Role of key components in the Nimrod-G architecture • Client or User Station acts as a user-interface for controlling and supervising an experiment under consideration. • Parametric Engine acts as a persistent job control agent and is the central component from where the whole experiment is managed and maintained. • Scheduler is responsible for resource discovery resource selection, and job assignment. • Dispatcher primarily initiates the execution of a task on the selected resource as per the scheduler’s instruction. • Job-Wrapper is responsible for staging of application tasks and data; starting execution of the task on the assigned resource and sending results back to the parametric engine via dispatcher.

38. Nimrod-G (cont’) • Deadline and budget-constrained algorithms • Deadline-constrained cost minimization heuristic • Budget-constrained time minimization heuristic • Joint time and cost minimization

39. Nimrod-G (cont’) • Deadline and budget-constrained algorithms • Deadline-constrained cost minimization heuristic • Sort resources by increasing cost • For each resource in order, assign as many jobs as possible to the resource, without exceeding the deadline • Repeat all steps until all jobs are processed

40. Nimrod-G (cont’) • Deadline and budget-constrained algorithms • Budget-constrained time minimization heuristic • For each resource • calculate the next completion time for an assigned job • taking into account previously assigned jobs • Sort resources by next completion time • Assign one job to the first resource for which the cost per job is less than the remaining budget per job • Repeat all steps until all jobs are processed

41. Nimrod-G (cont’) • Deadline and budget-constrained algorithms • Joint time and cost minimization • Split resources by whether cost per job is less than budget per job • For the cheaper resources, assign jobs in inverse proportion to the job completion time • For the dearer resources, repeat all steps (with a recalculated budget per job) until all jobs are assigned; repeat all steps until all jobs are processed

42. References • [1] Jörn Altmann, Costas Courcoubetis, John Darlington, Jeremy Cohen, GridEcon – The Economic-Enhanced Next-Generation Internet, GECON 2007, Workshop on Grid Economics and Business Models, Springer LNCS, Rennes, France, August 2007. • [2] Rajkumar Buyya, David Abramson, Jonathan Giddy, and Heinz Stockinger, Economic Models for Resource Management and Scheduling in Grid Computing, Journal of Concurrency: Practice and Experience, Grid computing special issue 14/13-15, 2002, pp 1507 – 1542. • [3] Rajkumar Buyya, David Abramson, and Jonathan Giddy, Nimrod/G: An Architecture for a Resource Management and Scheduling System in a Global Computational Grid, Proceedings of the 4th International Conference and Exhibition on High Performance Computing in Asia-Pacific Region (HPC ASIA 2000), May 14-17, 2000, Beijing, China, IEEE CS Press, USA, 2000. • [4] Chris Kenyon, Grid Economics - Grid Value and Practical Realization, Slide of Zurich Research Lab.