presentazione di powerpoint€¦ · vc startup. a vc example. environment layout: • 72 computing...
TRANSCRIPT
• 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