Skip to content

Latest commit

 

History

History
executable file
·
182 lines (143 loc) · 6.67 KB

README.md

File metadata and controls

executable file
·
182 lines (143 loc) · 6.67 KB

Researchers Cloud

code-grade code-quality vulnerabilities-shield downloads-shield contributors-shield

A simulation library based on Cloudsim that aims to offer many methods to automate replication of various simulation scenarios.

Why not use Cloudsim directly?

I faced alot of difficulties during my masters studies and alot of time was consumed by the process of re-implementating the research papers' algorithm/s instead of focusing on the actual improvments that could be done on any algorithm. Therefore, i created this tool to help me replicate all of those various experimentation scenarios with few lines of code as possible and as scalable and flexible as i can make them.

Features Include:

  • Simplify process of creating, modifying and deleting cloudsim entities.
  • Create random attributes and automate enitity instantiation.
  • Control the distribution of created cloudlets over brokers, such as: Even or Random.
  • Integrate scheduling algorithms and control cloudsim process easily.
  • Packed with implementations of several famous cloud algorithms such as: Honey Bee and PSO.
  • Display experiment outcomes in form of tables, charts and log files.

Help me expand base of implemented cloud algorithms

Simplest example

Create a ReCloud instance to start configuring experiment fields, then launch instance through Recloud.launch():

// Create a ReCloud instance.
ReCloud recloud = new ReCloud();

// Create a datacenters/server.
recloud.servers().newServer().make();

// Create a host.
recloud.servers().newHost().make();

// Create a broker.
recloud.jobs().newBroker().make();

// Create a virtual machine.
recloud.jobs().newVm().make();

// Create a cloudlet (Task).
recloud.jobs().newTask().make();

// Add wanted simulations to experiment.
recloud.experiment().newSimulations(new CloudsimSimulation());

// Add goal for number of cloudlets created.
recloud.experiment().taskTargets(100);

// Launch.
ReCloud.launch(recloud);

Intermediate example

Configure each cloud entity separatly and run multiple scheduling algorithms with the same setup environment:

// Create a ReCloud instance.
ReCloud recloud = new ReCloud();

// Create and configure datacenters/server.
recloud.servers().newServer().name("Server").environment("x86", "Linux", "Xen").timeZone(10.0)
.secCost(3.0).memCost(0.05).storageCost(0.001).bwCost(0.0).intervals(0).clones(1).make();

// Create and configure host.
recloud.servers().newHost().on("Server").mips(177730).pes(6).ram(16000).bw(15000)
.storage(4000000).clones(2).make();

// Create and configure broker.
recloud.jobs().newBroker().name("koala").make();

// Create and configure virtual machine.
recloud.jobs().newVm().mips(9726).pes(1).ram2(9).bw(1000).image(10000).vmm("Xen").clones(5).make();

// Create and configure a cloudlet (Task).
recloud.jobs().newTask().randomStyle(RandomStyle.Fixed_Pace).length(10000, 20000).pes(1)
.filesize(1).outpusize(1).make();

// Add wanted simulations to experiment.
recloud.experiment().newSimulations(
 new CloudsimSimulation(), 
 new Bullet_CS(GunType.Magnum),
 new SJF_CS(), 
 new HoneyBee_CS(0.1)
);

// Add goal for number of cloudlets created.
recloud.experiment().taskTargets(100);

// Launch.
ReCloud.launch(recloud);

Custom example

The difficulty of more complex examples rise with the experiment parameters, say we need the following setup:

[01] 2 servers | server_1 & server_2.
[02] 2 hosts | host_1 on server_1 & server_2 | host_2 on server_1.
[03] 1 host | host_3 on server_2.
[04] 2 brokers | koala 1 & koala 2.
[05] 2 virtual machines | vm_1 on koala_1 | vm_2 on koala_1 & koala_2 with 3 identical clones.
[06] 1 virtual machine | vm_3 on koala_2 .
[07] 1 fixed task-type* cloudlets.
[08] 1 random task-type* cloudlets on koala_2.

task-types*: are cloudlets templates that will generate cloudsim cloudlets/tasks in runtime according to their setup.

And with this setup we need to configure the experiment as following:

[09] Create cloudlets and ditribute them over brokers randomly. [10] Test CloudsimSimulation, Bullet_Magnum and Particle Swarm Optimization algorithms. [11] Experiment with 100, 200, 300 tasks.

// Create a ReCloud instance.
ReCloud recloud = new ReCloud();

// [01]
recloud.servers().newServer().name("server_1").make();
recloud.servers().newServer().name("server_2").make();

// [02]
recloud.servers().newHost().on("server_1", "server_2").make();
recloud.servers().newHost().on("server_1").make();

// [03]
recloud.servers().newHost().on("server_2").make();

// [04]
recloud.jobs().newBroker().name("koala_1").make();
recloud.jobs().newBroker().name("koala_2").make();

// [05]
recloud.jobs().newVm().on("koala_1").make();
recloud.jobs().newVm().on("koala_1").clones(3).make();

// [06]
recloud.jobs().newVm().on("koala_2").make();

// [07]
recloud.jobs().newTask().length(10000).pes(1).filesize(1).outpusize(1).make();

// [08]
recloud.jobs().newTask().on("koala_2").length(10000, 20000).pes(1).filesize(1).outpusize(1).make();

// [09]
recloud.jobs().taskSplit(TasksSplit.Random);

// [10]
recloud.experiment().newSimulations(
    new CloudsimSimulation(), 
    new Bullet_CS(GunType.Magnum),
    new PSO_CS(100, 1000, 0.1, 0.9, 1.49445, 1.49445, 5, Inertia.LDIW_1, Position.Sigmoid)
);

// [11]
recloud.experiment().taskTargets(100, 200, 300);

// Launch.
ReCloud.launch(recloud);