在集群和云上运行并行优化 - Concepts NREC解决方案.pptxVIP

Running Parallel Optimization on Clusters and Clouds;Overview of IOSO (Oleg) Overview of Cluster/Parallel Capabilities (Oleg) Overview of Amazon Offering, Start Demo (EPC) Axial Machine Examples (Oleg) Pump Optimization Example (EPC) Overview of Demo B2B (Oleg) Live Demonstration Progress (EPC) Questions ;Overview of IOSO(Oleg);Optimization in Integrated Agile Engineering Design System;Overview of IOSO;Parallelized IOSO;Overview of parallel CLUSTER(oleg);Loosely Defined “Cluster” for parallelized optimization;Overview of amazon ec2 offering (edward);Blade to Blade CFD 8114 Nodes – 4 Grid Blocks – MPI = 4 600 Iterations Laptop 4 Core Run Time: 72 sec. EC2 2 x 2 Core Master/Slave – MPI = 2 (Each) 2 x c1.medium Run Time: 100 sec. Cloud/VM Overhead Factor: ~1.4 $0.285/hr. per instance 64 Instances ~$18.25/hr. ;Scaling of IOSO Parallel on Amazon EC2;Virtual Private Cloud;Elastic IP and Network Interface;Running EC2 Instances;Stored Amazon Machine Images;Instance Setup – Choose AMI;Instance Setup – Choose Instance Type;Instance Setup – Choose VPC Subnet;Instance Setup – Set Private IP Address;Instance Setup – Choose Security Group;Instance Setup – Launch!;Axial Machine Examples;Optimization Examples;Optimization Examples;Optimization Examples;Optimization Examples;Optimization Examples;Optimization Examples;Pump impeller optimization with ioso (EPC);Base Case CFD Grid;Base Case CFD;Single Stage, Water, 300 gal/min 3D CFD Calculations at 100%, 60%, 40% DF Maximize EFF TT and Head Rise at Design Flow Rate Constrain Head Rise, EFF TS/TT at Off Design Flow Rates to Meet or Exceed Base Case Values Constrain CFD to Be Complete and Converged Screen for Improbable Efficiency and Head Rise IOSO PM (Parallel Cluster);Quantity;Quantity;~491 Runs 37.5 min. X 3 Flow Rates X 491 ~ 38 days CPU Using Cluster (15) ~ 2.5 days 2 Pareto Points 145 Feasible Solutions Improvements 14 in EFF TT 22 in Head Rise;Run1 - History;Run 1 - Efficiency Improvement;Run 1 - Head Rise Improvement;Run 1