principles of computing resources planning in cloud-based problem solving environment
TRANSCRIPT
Principles of planning the
computing resources in cloud-based
problem solving environment
K. Borodulin and G. RadchenkoSouth Ural State University
The reported paper was supported by RFBR, research project No. 15-29-07959.
Problem definition• Modern problems of computational science are
characterised by high demands on provided computing resources and complex computing tasks structure, which can be defined as a workflow.
• Also, problems of this type are characterised by usage of multivariant calculations computing task is run hundreds or thousands of times with different variations of the input parameters.
2
3
Example workflow for mixer simulation
Design Modeler
CFX-Mesh CFX-Pre
CFX-Solver
CFX-Post
Creation or correction of a
geometric model
Creation or correction of a mesh
Creation or correction of problem
definition
Problem Simulation in CFX-Solver
Visualization in CFX-Post
Values of the optimization
criteria are not satisfactory
The accuracy of the calculation is not satisfactory
4
Problem-solving environment
• Problem-solving environment is a program system that warps and provides a problem-oriented access to computational resources to solve a specific class e-Science problems
• This limitation would allow using a domain-specific information of task to predict a computational characteristics of the task in planning and scheduling workflow’s execution, inflating the efficiency of available computing resources’ consumption in cloud computing system.
03.10.16
5
Scheduler
Cloud Platform
DiVTB Web Interface
A Driver
Simulation Results
EngineerDistributed Virtual Test Bed(DiVTB ) includes an interface for a problem statement; a driver (a set of software tools
enabling the use of cloud resources for virtual experiment);
a set of services (a set of images of virtual machines)
a set of computing resources (a cloud computing environment)
Distributed Virtual Test Bed
6
Goal and tasksThe aim of the paper is to describe the principles of computing resources planning in Cloud-based Problem Solving Environment. To gain the aim of paper:• Analyze related solutions for the planning of
execution of problem-solving workflow’s. • Define a structure of cloud system for problem-
solving environment’s deployment. • Describe a scheme of an approach for the
computing resources planning in Cloud-based problem-solving environment.
03.10.16
7
Scheduling methods in workflow systems
• Pandey, S., Wu, L., Guru, S.M., Buyya, R.: A particle swarm optimization-based heuristic for scheduling workflow applications in cloud computing environments. In: Proceedings - International Conference on Advanced Information Networking and Applications, AINA. pp. 400407 (2010).
do not use information about previous executions.• Sokolinsky, L.B., Shamakina, A. V.: Methods of resource
management in problem-oriented computing environment. Program. Comput. Softw. 42, 17–26 (2016).
• Nepovinnykh, E.A., Radchenko, G.I.: Problem-Oriented Scheduling of Cloud Applications : PO-HEFT Algorithm Case Study. 2016 39th Int. Conv. Inf. Commun. Technol. Electron. Microelectron. MIPRO 2016 - Proc. 196–201 (2016).
scheduling the workflow actions’ execution - when the task has to run to provide a minimal makespan, for example.
03.10.16
8
Model of cloud-based problem-solving
environment• Components of Cloud-based
problem-solving environment• Input data for planning the
computing resources• Levels of planning the computing
resources in cloud-based problem solving environment
03.10.16
9
Components of cloud system
Base image• OS• Middleware
• Initialization service
• agent of remote task execution
• Agent of monitoring system
Software• Application
or software for task’s execution
• Service of software execution
Service
03.10.16
10
Virtual machine
Mem
CPU
Storage
Net
VMx
Service 1
NodeCPU Mem Ne
t
Local StorageVM1
VM2
Network Storage
03.10.16
11
Input data• Executable workflow• A set of the domain-
specific arguments’ values
• QoSo Makespano A number of the computing
resourceso Maximum of the cost
• Path of Input files• Path of result files
03.10.16
12
VM1
VM2
VM3
VM4
Vm6
VM5 VM7
T1T2
Node Node
VM1
Vm6
VM3VM2 VM4
VM7VM5T1
T2
Workflow level
Virtual machine Level
Computing nodes level
Servic
e
T1
T2
Service level
03.10.16
13
Workflow levelThe workflow layer implements the transformation of the abstract workflow into the executable job. • The data sources of the input
parameters are being connected with the certain tasks and sub-flows of the workflow during the transformation.
• The abstract workflow is executable if input arguments of any task are independent of the result of another task’s execution.
03.10.16
Creation or correction of a
geometric model
Creation or correction of a mesh
Creation or correction of problem
definition
Problem Simulation in CFX-Solver
Visualization in CFX-Post
Values of the optimization
criteria are not satisfactory
The accuracy of the calculation is not satisfactory
14
Service levelThe Service layer provides assignment of particular services to the required computing resources for any task in the workflow. The Workflow predictor sends to the Workflow planner the following prediction information: • time of the task execution (on the 1 computing core);• the amount of main memory, needed for the task
execution;• maximum task scaling (how much cores can be
provided to the task);• the amount of the result data;• prediction accuracy for each value.
03.10.16
15
Virtual machine levelThis layer performs the instances selection, using the prediction of computing resources required for the certain task execution, but • Workflow executor can allocate another set of
resources for QoS satisfaction, • If the prediction accuracy is low, i.e. most likely
prediction if false, then executor choose the type of virtual machine which is default for the certain service.
03.10.16
16
Computing node levelThe Computing node planning layer maps virtual machines’ onto computing nodes on the basis of a virtual machine computing resources and a volume of node’s local storage. At this layer, planner tends to place related virtual machine from the workflow (which on this layer are presented as Task-to-VM list on the same node to reduce the amount of data are transferred between the computing nodes.
03.10.16
17
Tasks that need to be addressed
Future work:• Development of workflow applications planning
algorithm with effective virtual machines allocation and possibility of dynamically adjustment of the execution plan of the application.
• Development of a database to support an estimation of execution characteristics of calculation tasks.
• Development of a model of workflow execution in cloud computing environment.
• Development of an experimental “Problem-oriented Scheduler” system
03.10.16
18
Thanks for your attention
03.10.16