Modelling LIT Cloud Infrastructure at JINR and Evaluating the Model Reporter: Vagram Airiian Master’s student, Dubna State University GRID’16. 05.07.2016
Motivation To create a model comprising LIT Cloud Infrastructure allowing to scrutinize its functioning to develop and apply a virtual machines migration scheme for freeing a maximum amount of computing resources in order to reutilize them.
Problem Model Current configuration 30 x: 4 CPU cores, 8 GB RAM, 250 GB HDD 3 x: 2 CPU cores, 4 GB RAM, 800 GB HDD 1 x: 4 CPU cores, 8 GB RAM, 2330 GB HDD 4 x: 12 CPU cores, 36 GB RAM, 1000 GB HDD 1 x: 24 CPU cores, 24 GB RAM, 1000 GB HDD Total: 202 CPU cores, 428 GB RAM, 17230 GB HDD Model
Frameworks Comparison CloudSim GreenCloud iCanCloud SimGrid Platform any NS2 OMNET, MPI Language Java C++/OTcl C++ C, Lua, Java, Ruby Availability Open Source Graphical Mode Limited (CloudAnalyst) Limited (Nam) Full Limited Physical Models None (plugin) Parallel Experiments No Yes Energy Consumption VM Live Migration
Frameworks Comparison (2) CloudSim GreenCloud SimGrid Language Java C++/OTcl C, Lua, Java, Ruby Distribution Source code Source code / prebuilt VM Source code / prebuilt binary Stable Release 02.05.2013 13.02.2016 13.10.2015 Latest Release 17.05.2016 Flexibility Some Full
SimGrid http://simgrid.gforge.inria.fr/documentation.php SimGrid Functional Organization: MSG: user-friendly syntaxic sugar Simix: processes, synchronization SURF: resources usage interface LMM: linear systems solver http://simgrid.gforge.inria.fr/documentation.php
SimGrid Example http://simgrid.gforge.inria.fr/documentation.php Platform description: public class Simulation extends Process {} Main() routine: org.simgrid.msg.Process http://simgrid.gforge.inria.fr/documentation.php
Simplified Model of Cloud Model configuration 5 host types 4 HDD and 4 RAM types 39 hosts combined 1 router connecting all hosts 1 link type – 1 Gbps 39 routes (780 auxiliary routes)
Migration Algorithm Pre-copy algorithm (SimGrid’s default)
Model Evaluation Migration game
Bin Packing Problem Given: n objects need to be placed in bins of capacity V each. Object i requires Vi units of bin capacity. Object: determine the minimum number of bins needed to accommodate all n objects.
To Do Extend the model with a definition of host capacity and develop our own migration scheme based on bin packing algorithms in respect of long-time statistics. Investigate OpenVZ and KVM usage in the cloud and apply the developed scheme to them.
Thank You for Attention!
References Plattner René. A Comprehensive Grid and Network Simulation Tool for Workflow based Applications – Master’s Thesis in CS, Distributed & Parallel Systems Group, Institute of Computer Science, the University of Innsbruck, 2007. William Voorsluys, James Broberg, Srikumar Venugopal, Rajkumar Buyya. Cost of Virtual Machine Live Migration in Clouds: A Performance Evaluation – arXiv:1109.4974 [cs.DC], 9 December 2011. Anton Beloglazov, Jemal Abawajy, Rajkumar Buyya. Energy-aware resource allocation heuristics for efficient management of data centers for Cloud computing – Future Generation Computer Systems, Vol. 28, Issue 5, May 2012, Pages 755–768. Anton Beloglazov. Energy-Efficient Management of Virtual Machines in Data Centers for Cloud Computing – PhD Thesis in CS, Department of Computing and Information Systems, the University of Melbourne, 2013. Astrikov D.Yu., Kuz'min D.A., Panasyuk A.I. Modelirovanie sistemy planirovaniya raspredelennogo vysokoproizvoditel'nogo vychislitel'nogo kompleksa [Simulation of a scheduling system of the distributed high-performance computing system]. Doklady Akademii Nauk Vysshei Shkoly Rossiiskoi Federatsii – Proceedings of the Russian Higher School Academy of Sciences, 2014, no. 2-3 (23-24), pp. 34-41. P.A. Mihailov, G.I. Radchenko. Modeling and Performance Evaluation of Cloud Systems – Bulletin of the South Ural State University, 2014, Vol. 3, no. 3, pp. 109–123 Xin Lu and Zhuanzhuan Zhang, A Virtual Machine Dynamic Migration Scheduling Model Based on MBFD Algorithm – International Journal of Computer Theory and Engineering, Vol. 7, No. 4, August 2015