Transcript TNC 2005

Enabling Grids for E-sciencE
TNC 2005
Networking activities in EGEE
Mathieu Goutelle (CNRS UREC, France)
EGEE-SA2 activity
Poznan (PL), 2005-06-07
www.eu-egee.org
INFSO-RI-508833
EGEE in a nutshell
Enabling Grids for E-sciencE
• A 2-year European project within a 4-year program (2004-2008),
70 partners (850 people involved) across 27 countries,
about €32M for the first two years,
• Federate regional and national grid in a seamless infrastructure:
 More than 120 sites across the world,
 More than 12 000 CPUs,
 More than 5 PB storage.
• A new middleware ― gLite ― based
on many of the preceding ones,
• Applications:





High Energy Physics (LHC),
Biomedical applications,
Earth Science and Geophysics,
Computational Chemistry,
…
• With the network in between!
INFSO-RI-508833
TNC 2005 - Poznan (PL), 2005-05-07
2
Computing Resources – Apr. 2005
Enabling Grids for E-sciencE
Country providing resources
Country anticipating joining
In LCG-2:
 131 sites, 30 countries
 more than 12,000 CPU
 about 5 PB storage
Includes non-EGEE sites:
• 9 countries
• 20 sites
INFSO-RI-508833
TNC 2005 - Poznan (PL), 2005-05-07
3
EGEE Networking Activities
Enabling Grids for E-sciencE
• SA2: Network Resource Provision
– Technical Network Liaison Committee,
– Operational interface with NRENs,
– SLAs installation between EGEE  Network (Geant & NRENs).
• JRA4: Network Services Development
– Network Performance Monitoring,
– Bandwidth Allocation and Reservation,
– IPv6 Uptake.
Building important working relation between EGEE and
the network providers (Geant & NRENs)
INFSO-RI-508833
TNC 2005 - Poznan (PL), 2005-05-07
4
Our approach
Enabling Grids for E-sciencE
• Network must be viewed as a class of Grid resource:
– Like computing and storage resources,
– Implies a network resource provisioning architecture,
– Problem to manage a resource outside the EGEE world!
• This resource depends on the network services provided by the
network providers (NRENs, Geant):
– Current services: Premium IP (even not in all NRENs),
– We need to anticipate on the future GN2 services:
 Currently in the definition process,
 GN2 project started 6 months later than EGEE.
• Network Activities work in a long term perspective:
– Most of the tasks started from scratch: Operational interface, SLA,
NPM, BAR…
 Except the monitoring tools coming from past projects (Datagrid WP7).
INFSO-RI-508833
TNC 2005 - Poznan (PL), 2005-05-07
5
Operational interface
Enabling Grids for E-sciencE
• Define the interactions between the Grid User support
and the NREN NOCs:
 SLA management (processing, installation, monitoring),
 Trouble management (reporting, interactions).
– Definition of procedures and information flows between NOCs
and EGEE:
 Need of a consistent view of the “EGEE network”,
 Need to evaluate its reliability and the quality of the services,
 Will not replace the current interactions between a resource centre
and its NREN.
• Progressive integration into the existing Global Grid
User Support (GGUS, http://www.ggus.org/):
– Single entry point for the support to the Grid users,
– Difficulties remain (trouble tickets normalization, languages…).
INFSO-RI-508833
TNC 2005 - Poznan (PL), 2005-05-07
6
GEANT/NRENs notification handling
Enabling Grids for E-sciencE
E-mail notifications
NREN
GEANT
NREN
NREN
Handling
Procedure
ENOC
Filtering
Network related
TTS
INFSO-RI-508833
Create an
assigned ticket
TNC 2005 - Poznan (PL), 2005-05-07
7
Troubleshooting workflow
Enabling Grids for E-sciencE
Problem
problem
assignment
essiment
ENOC
TTM
Feedback
Feedback
Problem
Problem
signalisation
signalling
NOC B
USER
PERT
Problem
resolution
failure
RC
NREN A
GEANT
NREN B
RC
TTM: Trouble Ticket manager
ENOC: EGEE Network Operation Center
INFSO-RI-508833
TNC 2005 - Poznan (PL), 2005-05-07
8
SLA
Enabling Grids for E-sciencE
• SLA definition:
– Based on previous work and answers inside EGEE and from
GN2 to some open issues (procedures, demarcation point…),
– Definition in cooperation with GN2.
• EGEE end-to-end SLA template:
– SLA between the border of the NRENs cloud (border-to-border
SLA),
– Difficulty to accommodate and take into account the “last mile”.
INFSO-RI-508833
TNC 2005 - Poznan (PL), 2005-05-07
9
SLA institution scenarios
Enabling Grids for E-sciencE
• All domains involved in network services provisioning
to EGEE as part of the existing network infrastructure
hierarchy have to be categorized as:
– Compliant with the Premium IP service
– Supportive of the Premium IP service
– Indifferent to the Premium IP service
INFSO-RI-508833
TNC 2005 - Poznan (PL), 2005-05-07
10
QoS experiment
Enabling Grids for E-sciencE
• Network service use case for an application:
– Choice of a particular “interactive” application,
– Aim:
 Better knowledge of the SLA processing,
 Better specifications of the applications requirements,
 Precision about procedures and identified issues.
• Status:
– Validation on a local platform:
 Validation of middleware modifications (packets marking),
 Validation of the experimental protocol.
– Tests on long distance networks:
 First in the same administrative domain (Renater),
 Then, tests involving three different networks,
 Involved networks: Renater, GEANT, GRNET.
INFSO-RI-508833
TNC 2005 - Poznan (PL), 2005-05-07
11
The picture
Enabling Grids for E-sciencE
Lyon, France
Tag IP packets with an “inoffensive” (< 40)
DSCP value (16 in the experiment).
GEANT
GRNET
Renater
The first router remarks packets
with the standard DSCP value for
PIP (46).
Athens, Greece
INFSO-RI-508833
TNC 2005 - Poznan (PL), 2005-05-07
12
EGEE Networking Activities
Enabling Grids for E-sciencE
• SA2: Network Resource Provision
– Technical Network Liaison Committee,
– Operational interface with NRENs,
– SLAs installation between EGEE  Network (Geant & NRENs).
• JRA4: Network Services Development
– Network Performance Monitoring,
– Bandwidth Allocation and Reservation,
– IPv6 uptake.
INFSO-RI-508833
TNC 2005 - Poznan (PL), 2005-05-07
13
Network Performance Monitoring
Enabling Grids for E-sciencE
• JRA4 approach:
 Standardization of access to network performance monitoring
across different domains and frameworks,
 GGF NM-WG recommendation is the selected basis for
standardisation.
• Purpose:
– Provide to Grid operations networking information for monitoring
and troubleshooting,
– SLA monitoring,
– Information publication in the Grid Information System.
• Potential user : middleware, end-user, Grid operations
• Status:
– Prototype ready (demo during the last GGF),
– Security (on-going work inside EGEE),
– Work on the interactions with the middleware.
INFSO-RI-508833
TNC 2005 - Poznan (PL), 2005-05-07
14
Network Performance Monitoring
Enabling Grids for E-sciencE
Some Client
NM-WG
NM-WG
End Site
EDG WP7
End Site
Home grown
INFSO-RI-508833
NM-WG
NM-WG
NM-WG
Backbone
Perfmonit
Backbone
PiPEs
Backbone
GN2
TNC 2005 - Poznan (PL), 2005-05-07
15
Network Performance Monitoring
Enabling Grids for E-sciencE
Diagnostic Client
NM-WG
JRA4 NPM Mediator
NM-WG
NM-WG
End Site
EDG WP7
End Site
Home grown
INFSO-RI-508833
NM-WG
NM-WG
NM-WG
Backbone
Perfmonit
Backbone
PiPEs
Backbone
GN2
TNC 2005 - Poznan (PL), 2005-05-07
16
BAR
Enabling Grids for E-sciencE
• Purpose:
– Provide Grid users the ability to use network services
(reservations),
– Interface with the mechanisms that will be available in the
network,
– Will first propose the Premium IP service.
• Status:
– Step-by-step integration with the network services:
 Static configuration between 2 backbone sites ― june 2005,
 Investigation of the “last mile” problem ― oct. 2005,
 Pilot reservation system in limited number of site ― jan. 2006.
– Need of a strong participation of NRENs (GN2).
INFSO-RI-508833
TNC 2005 - Poznan (PL), 2005-05-07
17
BAR Architecture
Enabling Grids for E-sciencE
Site 1
Network 1
Network 2
Network 3
Site 2
HLM
BAR
BAR
EGEE
Network
L-NSAP
L-Network
INFSO-RI-508833
NSAP
NSAP
Core Network
NSAP
L-NSAP
L-Network
TNC 2005 - Poznan (PL), 2005-05-07
18
Reservation Architecture
Enabling Grids for E-sciencE
• Integration of network resource in a more complex
architecture:
– Essential to meet the applications requirements,
– Convergence between the allocation and reservation
architecture of the EGEE middleware and the reservation
architecture developed by GN2,
– Joint work of GN2 and EGEE.
• Status:
– GN2 architecture is to be defined:
 Based on inter-domain SLAs,
 A still very manual prototype is foreseen due beginning of next year.
– Preparatory work inside EGEE :
 Convergence to a common architecture,
 Consistency between the two architectures.
INFSO-RI-508833
TNC 2005 - Poznan (PL), 2005-05-07
19
Conclusion
Enabling Grids for E-sciencE
• Grids are only one user of network services (even if big
and federated) among many other potential ones:
– Special requirements for data transfer (lightpath?),
– For other requirements (interactive messages and streams),
need of a service to request from the network the use of services
(shared with the other users).
• Good progress in terms of SLAs definitions: will the
NRENs support our model?
• The operational interface is needed to build a reliable
computing architecture (again very challenging),
• Further step remains towards the fulfilment of
applications requirements in term of network
resources,
• Good working relations with NRENs and GN2.
INFSO-RI-508833
TNC 2005 - Poznan (PL), 2005-05-07
20
Enabling Grids for E-sciencE
Thank you!
General website: http://www.eu-egee.org/
JRA4 website: http://egee-jra4.web.cern.ch/egee-jra4/
SA2 website: http://egee-sa2.web.cern.ch/egee-sa2/
INFSO-RI-508833
TNC 2005 - Poznan (PL), 2005-05-07
21