Presentazione di PowerPoint€¦ · VC Startup. A VC Example. Environment layout: • 72 Computing...

Vida

Virtual Images Deoyment Architecture

Paolo Anedda, CRS4paolo.anedda@crs4.it

• Cybersar

• Introduction to SOC

• VIDA Architecture

• Tests Evaluation

• Future works

• Q&A

Summary

• A Cyber-infrastructure for research in Sardinia based on high speed networks interconnecting the island research communities and computational facilities.

• Cybersar infrastructure is designed so that it can support innovative computational applications while providing an experimental platform for research on application driven orchestration of computing and network resources.

• Cybersar is managed by Cosmolab, a consortium that includes all the main public and private organizations in Sardinia interested in High Performace Computing and Networks and its applications:

CRS4 (Center for Advanced Studies, Research and Development in Sardinia)

INAF (Istituto Nazionale di Astro Fisica)

INFN (Istituto Nazionale di Fisica Nucleare)

University of Cagliari

University of Sassari

Nice (company specialized in Grid Computing solutions)

Tiscali (European telecomunication company)

Cybersar

• Infrastructure for supporting Scientific Applications:– Massive computation in Astrophysics.

– Ensemble forecast system of “extreme” meteorological events.

– Wavefield migration and montecarlo imaging of deep underground structures.

– High Energy Physics.

Cybersar

• A SOA based virtual images deployment architecture.

• Provides an efficient environment for Virtual Computational Clusters:

– Manageability.

– Isolation.

– Security.

• Supports resources control via web services.

• Xen VM support.

• Based on Open Source Software.

VIDA

• There has been an increase in interest recently towards "Service Oriented'' Computing.

• Services are often seen as a natural progression from component based software development.

• Scientific institutions are following the market trends.

Introduction

• Service-Oriented Architecture

• Service-Orientation Design Paradigm

• Service Composition

• Service Inventory

• Service-Oriented Solution Logic

• Services

SOC Elements

• A typical Web service is comprised of the following:

– A physically decoupled technical service contract.

– A body of programming logic.

– Message processing logic.

• The Web Services platform is defined through a number of industry standards:

– WSDL: Web Service Description Language.

– SOAP: Simple Object Access Protocol.

– UDDI: Universal Description Discovery and Integration.

Web Service

• Automatic deployment of virtual images.

• Resources scheduling.

• Separation between the management of physical resources and virtual resources.

• Support for different model of virtual clusters.

• Jolie-based Workflows implementation.

• WSDL-based services definition.

• Implementation based on OpenSource software.

VIDA Main Features

• The lower layer provides the access to the devices functionalities.

• Simple services are composed to form complex services.

• Applications processes are designed through the use of complex services.

VIDA Architecture Stack

• The design process is divided into four logical layers:

– Devices Infrastructure.

– Simple Services.

– Complex Services.

– Applications.

Architecture

Control Plane Services Composition

Machine management

● Virtual Resource Manager:– Schedules virtual resources

allocation.

● Virtual Machine Manager:– Creates the Virtual Clusters.

● Vlan Manager:– Creates and manages the Vlans.

● VMH:– Creates and manages the VMs.

Virtual Machine Management

• VMH (Virtual Machine Host) is the service responsible for the creation and control of the virtual images inside a real host.

• Every operation is mapped into a web service method.

• VMH communicates with the hypervisor via the Xen API.

• The lifecycle of a virtual host is characterized by five states.

• Each state transition is triggered by a VMH method invocation.

• Migration is not yet supported.

VMH

• Devices communication based on WS proxies.

• Devices control based on an Object model API.

• Support for multiple vendor.

Network management

VC creation workflow

VC Startup

A VC Example

Environment layout:

• 72 Computing nodes:– CPU: 2 x Dual-Core 2.6GHz

AMD Opteron.

– RAM: 8Gbyte.

– Disk: 2 x 250Gbyte SATA.

• 4 Computational clusters.

• 3 Nic per node:– 2 x 1 Gbyte Data exchange.

– 1 IPMI.

Tests evaluation

• Transfer based on Bittorrent protocol.

• Average transfer speed per client.

• Linear scalability.

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

0.00

5.00

10.00

15.00

20.00

25.00

30.00

35.00

40.00

45.00

50.00

SpeedMax ValMin Val

Clients

Sp

eed

(MB

yte/

sec)

Preliminary Tests Results

• Transfer based on Bittorrent protocol.

• 72 nodes cluster.

• 4.5 GByte image size.

• Average deployment time: 323.49 s

• Average trasfer time: 248.47 s

12

34

56

78

910

1112

1314

1516

1718

1920

2122

2324

2526

2728

2930

3132

3334

3536

3738

3940

4142

4344

4546

4748

4950

5152

5354

5556

5758

5960

6162

6364

6566

6768

6970

7172

73

0.00

50.00

100.00

150.00

200.00

250.00

300.00

350.00

400.00

450.00

Deployment time

Transfert time

Nodes

Tim

e

Tests Results

• Integration of different virtualization middleware (ex. OpenNebula).

• Integration of external scheduler.

• Support for different remote controlling chassis systems.

• Performances improvement.

Future works

• Xen: http://www.xen.org

• Jolie: http://www.jolie-lang.org

• CRS4: http://www.crs4.it

• Cybersar: http://www.cybersar.it

References