HEAnet Conference 2006

lilactruckInternet and Web Development

Dec 4, 2013 (3 years and 8 months ago)

73 views

HEAnet Conference 2006

John Walsh

Grid
-
Ireland Grid Manager

Trinity College Dublin

The Grid Computing
Infrastructure in Ireland and
Abroad

What is a Grid?


Collection of shared resources + infrastructure


Cooperating across multiple administrative domains


Secured with PKI (AAA)


Group Collaboration (Virtual Organisations)


Compute and/or Data orientated


Batch queue, Storage and Worker nodes


Infrastructure to support


Authentication and Authorisation (VOMS, MYPROXY)


Recognised Trust Authorities (Asia, Europe, US)


Resource discovery (BDII, LFC catalogs)


Current Grid projects are the drivers for massive
distributed data access over the next few years

Grid
-
Ireland: Infrastructure


NRENs+Grid+HPC ==
e
-
Infrastructure


Grid
-
Ireland:


The national computational Grid for Ireland


UK:


UK e
-
Science: NGS


GridPP: partners in EGEE


LCG/EGEE:


200+ sites and 31,124 CPUs worldwide


32 countries, 11 federations


TCD is Regional Operations Centre for Ireland


int.eu.grid:


European Interactive Grid


13 countries, no federations

Grid
-
Ireland: Infrastructure

26
-
OCT
-
06:


200+ sites


31,124 CPUs

Grid
-
Ireland: Infrastructure

Central

services

Sites

Clusters

Grid
-
Ireland: Infrastructure

Grid
-
Ireland

Central

services

Sites

Clusters

Gridnm (NM)


Gridui (UI)


Gridmon (test WN)

Gridstore (SE)

Gridgate (CE)


Gridinstall (Quattor)

Network switch

UPS




Gridfw (firewall)

Gateway

Cluster

Grid
-
Ireland: Infrastructure

Installation server



Gridnm (NM)


Gridui (UI)


Gridmon (test WN)

Gridstore (SE)

Gridgate (CE)


Gridinstall (Quattor)

Network switch

UPS




Gridfw (firewall)

Grid
-
Ireland: Infrastructure

Computing element



Gridnm (NM)


Gridui (UI)


Gridmon (test WN)

Gridstore (SE)

Gridgate (CE)


Gridinstall (Quattor)

Network switch

UPS




Gridfw (firewall)

Grid
-
Ireland: Infrastructure

Storage element



Gridnm (NM)


Gridui (UI)


Gridmon (test WN)

Gridstore (SE)

Gridgate (CE)


Gridinstall (Quattor)

Network switch

UPS




Gridfw (firewall)

Grid
-
Ireland: Infrastructure

Test worker node



Gridnm (NM)


Gridui (UI)


Gridmon (test WN)

Gridstore (SE)

Gridgate (CE)


Gridinstall (Quattor)

Network switch

UPS




Gridfw (firewall)

Grid
-
Ireland: Infrastructure

User interface



Gridnm (NM)


Gridui (UI)


Gridmon (test WN)

Gridstore (SE)

Gridgate (CE)


Gridinstall (Quattor)

Network switch

UPS




Gridfw (firewall)

Grid
-
Ireland: Infrastructure

Network monitor:

Not currently installed



Gridnm (NM)


Gridui (UI)


Gridmon (test WN)

Gridstore (SE)

Gridgate (CE)


Gridinstall (Quattor)

Network switch

UPS




Gridfw (firewall)

Grid
-
Ireland: Infrastructure

Cluster(s):

Managed by local admins


OpsCentre supports integration


Various config & install options

Gridnm (NM)


Gridui (UI)


Gridmon (test WN)

Gridstore (SE)

Gridgate (CE)


Gridinstall (Quattor)

Network switch

UPS




Gridfw (firewall)

Grid
-
Ireland: Infrastructure

Grid Gateway:

All Virtual machines


All run on 1 physical machine


Remotely managed by OpsCentre


OpsCentre:


Management of Grid
-
Ireland


Testing, porting, customisation


Deployment to remote sites


Remote management of sites


Monitoring of Grid services


National services (e.g. CA)


Dissemination and training


Grid courses and e
-
Learning (soon)


Links to EU Grids:


EGEE
-
II


int.eu.grid

OpsCentre

Grid
-
Ireland: Operations

Grid
-
Ireland: Operations

Transactional Deployment

Geoff Quigley, TCD

Typical Grid service
software stack.


LCG M/W stack
depends partially
on Globus stack


Grid
-
Ireland: Operations

Grid
-
Ireland: Operations

Remote Site

OpsCentre

Service monitoring

15,000 service checks per day

Keith Rochford, TCD

Grid
-
Ireland: Operations


National Trouble Ticketing:


HEAnet tickets
-
>
grid
-
ireland
-
mgr@cs.tcd.ie

(people)


Grid
-
Ireland alerts
-
>
grid
-
ireland
-
alert@cs.tcd.ie

(people)


Grid
-
Ireland tickets
-
>
grid
-
ireland
-
help@cs.tcd.ie

(OpsCentre RT)


Global Trouble Ticketing:


HEAnet tickets
-
> Global Grid User Support (GGUS)


Selected OpsCentre RT tickets <
--
> GGUS

Grid
-
Ireland: Intrusions

R
-
GMA

Secondary

Producer

Alert Archive

R
-
GMA

API

Snort

OpsCentre

Site A

R
-
GMA

API

Snort

Site B

R
-
GMA

API

Snort

Site C

R
-
GMA

API

Site D

Snort

GSMI Alert Layer

Stuart Kenny, TCD

Grid
-
Ireland: Intrusions

GSMI Analysis Layer

R
-
GMA

Secondary

Producer

Alert Archive

GOC

Alert

Analyser

Alert

Analyser

Alert

Analyser

Grid
-
Wide

Alerts

Correlation

Engine

Low / med priority
Grid

alerts

high

priority
Grid

alerts

OpsCentre

Stuart Kenny, TCD

Grid
-
Ireland: Intrusions


Deployed on Grid
-
Ireland


since July 2005


18 Irish sites


~500,000 alerts so far


Being deployed on int.eu.grid


13 EU sites


Next:


Tripwire sensors


Active Security

4 week period: 25,378

Stuart Kenny, TCD


Certification Authorities (CAs) issue digital certificates:


Grid CAs are recognized internationally:

http://www.gridpma.org/


e.g. Grid
-
Ireland CA:
http://www.grid.ie/certauth.html


Globus GSI software


basis of most production grid security


NRENs are investigating pan
-
EU federated identity


HEAnet and Grid
-
Ireland are proactively tracking this


Primary issue: users must keep their digital certificate secure !!!


Grid
-
Ireland: Trust

Grid
-
Ireland: Applications


Astrophysics


Computational Chemistry


Earth Sciences
-

Earth Observation, Solid Earth Physics


Geophysics


High Energy Physics


Life Sciences
-

Bioinformatics (Drug discovery, etc)


Medical Imaging


Multimedia


A J
-
band

extinction
map showing the
Chameleon dark clouds extracted from our
map created using the Grid
-
enabled program
-

G Murphy, D Froebrich, J Walsh (2005).

Grid
-
Ireland: Applications

Advances in Application Support



MPI


Currently not supported well


TCD involved in working group to rectify this


Will support multiple MPI implementations


TCD modified Resource Broker


Remote PBS


Secure Job submission to remote PBS servers


Standard PBS server with GSI
-
sshd + basic info publishers


Worker nodes have standard Glite WN M/W (e.g. tar ball)


Developed at TCD


Can be adapted to support other job managers


Currently beta (tested at TCD/UCD)


Questions?

Grid
-
Ireland