EEDC 34330 Intelligent Placement ofExecution Datacenters for InternetEnvironments for ServicesDistributedComputingMaster in Computer Architecture,Networks and Systems - CANS Homework number: 6 Umit Cavus Buyuksahin firstname.lastname@example.org
OUTLINE1. Introduction2. Example Datacenter3. Problem4. Placement of Datacenters5. Propose 5.1. Defining Framework 5.2. Formulation 5.3. Solving the problem6. Conclusion 2
Introduction Internet services reach the whole world. Millions of clients on the world. Demand high availabilityin short response time. Thus huge datacenters constructedaround the world They have many servers,cooling systems, energy power systems.. 3
Example - Datacenter Facebook - Prineville, Oregon USA – 147,000-square-foot facility – $200 million - $215 million.* http://www.oregonlive.com/business/index.ssf/2010/01/facebook_picks_prineville 4
Problem Clients ... widespreaded geographically ... demand high availablity ... in short response time Many servers requirement. Supplying Energy Cooling system Building and operating datacenters Green Energy 5
Problem Clients ... widespreaded geographically ... demand high availablity ... in short response time Many servers requirement. Supplying Energy Cooling system Building and operating datacenters Green Energy PLACEMENT OF DATACENTER !! 6
Placement of DatacenterDirect impact on ... Response time High availablity Mirrored Datacenters Closest one serves Capital and Operational Costs Land acquisition and building Bring network and electricy Electricity & Water Staff CO2 emmisions (indirect) 7
OUTLINE1. Introduction2. Example Datacenter3. Problem4. Placement of Datacenters5. Propose 5.1. Defining Framework 5.2. Formulation 5.3. Solving the problem6. Conclusion 8
ProposeDatacenter automation of palcement of data centers.. Selection and selection and automation,efficiently !! 9
Propose – Defining Framework Parameters Costs • CAPEX (Capital) bringing electricity and network land and construction power, backup, cooling equipment • OPEX (Operational) maintaince and administor electrcicity and water price Response Time • Latency & number of servers Consistency Delay • Latency from mirrored datacenters Availablity • #9 changes in each tier CO2 emissions 10
Propose – Formulation Subject to Minimizing CAPEX and OPEX Constraints Response times < MAX LATENCY , ∀ users Min consistency delay between 2 DCs < MAX DELAY Min system availability > MIN AVAILABILITY Output # of servers at each location Minimized cost 11
Propose – Solving Problem is ... non linear. ... not directly solvable by Linear Programming. Linear Programming (LP) for potential solution. Simulated Annealing (SA) for consiring neighborings. CA + LP for cost optimization. Quality of results compared with Brute solution. Tool is built ... automatic dacenter location selection ... new parameters and constraints can be added 12
Conclusion No other work for intelligent placement of datacenters. Contributions: A framework is proposed by defining parameters Based on parameters, optimization problem defined Proposed the most efficient and accurate solution approach A tool is built to automate location selection Experimental results shows Millions dollar are saved 14
A particular slide catching your eye?
Clipping is a handy way to collect important slides you want to go back to later.