G. Maron, INFN-GRID, Comm. I INFN, Cagliari, 13 settembre 2000 G. Maron, INFN-GRID, Comm. I INFN, Cagliari, 13 settembre 2000 1 1 LNL INFN-GRID-WP2.4: INFN-GRID-WP2.4: Computing Fabric Computing Fabric & Mass Storage & Mass Storage LNL LNL L. Berti L. Berti M. Biasotto M. Biasotto M. Gulmini M. Gulmini G. Maron G. Maron N. Toniolo N. Toniolo Padova Padova S. Balsamo S. Balsamo M. Bellato M. Bellato F. Costa F. Costa R. Ferrari R. Ferrari M. Michelotto M. Michelotto I. Saccarola I. Saccarola S. Ventura S. Ventura CNAF CNAF A. Chierici A. Chierici L. Dell’Agnello L. Dell’Agnello F. Giacomini F. Giacomini P. Matteuzzi P. Matteuzzi C. Vistoli C. Vistoli S. Zani S. Zani Lecce Lecce G. Aloisio G. Aloisio M. Cafaro M. Cafaro Z. Zzzz Z. Zzzz L. Depaolis L. Depaolis S. Campeggio S. Campeggio E. Fasanelli E. Fasanelli Torino Torino A. Forte A. Forte Genova Genova G. Chiola G. Chiola G. Ciaccio G. Ciaccio Roma 1 Roma 1 D. Anzellotti D. Anzellotti C. Battista C. Battista M. De Rossi M. De Rossi F. Marzano F. Marzano S. Falciano S. Falciano A. Spanu A. Spanu E. Valente E. Valente Bologna Bologna G.P. Siroli G.P. Siroli P. Mazzanti P. Mazzanti Catania Catania C. Rocca C. Rocca E. Cangiano E. Cangiano Tecnologo Tecnologo
30
Embed
LNL G. Maron, INFN-GRID, Comm. I INFN, Cagliari, 13 settembre 2000 1 INFN-GRID-WP2.4: Computing Fabric & Mass Storage LNL L. Berti M. Biasotto M. Gulmini.
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
G. Maron, INFN-GRID, Comm. I INFN, Cagliari, 13 settembre 2000G. Maron, INFN-GRID, Comm. I INFN, Cagliari, 13 settembre 2000 11
LNL
INFN-GRID-WP2.4:INFN-GRID-WP2.4:Computing Fabric Computing Fabric & Mass Storage& Mass Storage
G. Maron, INFN-GRID, Comm. I INFN, Cagliari, 13 settembre 2000G. Maron, INFN-GRID, Comm. I INFN, Cagliari, 13 settembre 2000 22
LNLTerminologyTerminology
PCPC
LANLAN
PC network PC network + midleware+ midleware
= PC Farm/Fabric= PC Farm/Fabric PC ClusterPC Cluster
LALANN
LALANN
LALANN
LALANN
LALANN
LALANN
PC Farm network PC Farm network + midleware+ midleware
= = GRIDGRID
G. Maron, INFN-GRID, Comm. I INFN, Cagliari, 13 settembre 2000G. Maron, INFN-GRID, Comm. I INFN, Cagliari, 13 settembre 2000 33
LNLWhy this WP?Why this WP?
Commodity components like PCs and LANs are mature to form Commodity components like PCs and LANs are mature to form inexpensive and powerful Computing Fabricinexpensive and powerful Computing Fabric
Computing Fabrics located in different sites are integrating to Computing Fabrics located in different sites are integrating to form a Computational/Data GRIDform a Computational/Data GRID
ButBut
– How to design a fabric of 1000s nodes balancing computing How to design a fabric of 1000s nodes balancing computing power and efficient storage access?power and efficient storage access?
– How to control and monitor the basic system components?How to control and monitor the basic system components?
– How to “publish” the monitored values to the GRID monitor How to “publish” the monitored values to the GRID monitor system?system?
G. Maron, INFN-GRID, Comm. I INFN, Cagliari, 13 settembre 2000G. Maron, INFN-GRID, Comm. I INFN, Cagliari, 13 settembre 2000 44
LNLComputing Fabric WPComputing Fabric WP
This WP wants to address the mentioned problems adding a technological This WP wants to address the mentioned problems adding a technological tracking task to follow and to test, with real use case, the evolution of the basic tracking task to follow and to test, with real use case, the evolution of the basic constituents of a fabric.constituents of a fabric.
Overall architecture and fabric setupOverall architecture and fabric setup LAN and SAN (System Area Network) technologiesLAN and SAN (System Area Network) technologies Communication protocols for high speed network fabricsCommunication protocols for high speed network fabrics Storage Systems Storage Systems Microprocessor TechnologyMicroprocessor Technology
– Fabric Management (DataGrid WP4)Fabric Management (DataGrid WP4) Configuration management and automatic software installationConfiguration management and automatic software installation System monitoringSystem monitoring Dynamic System PartitionDynamic System Partition Problem managementProblem management
G. Maron, INFN-GRID, Comm. I INFN, Cagliari, 13 settembre 2000G. Maron, INFN-GRID, Comm. I INFN, Cagliari, 13 settembre 2000 55
Konrad Zuse Zentrum (Berlin)Konrad Zuse Zentrum (Berlin)
Kirchhoff Institute (Heidelberg)Kirchhoff Institute (Heidelberg)
IN2P3 (Lyon)IN2P3 (Lyon)
INFNINFN
NikhefNikhef
RALRAL
G. Maron, INFN-GRID, Comm. I INFN, Cagliari, 13 settembre 2000G. Maron, INFN-GRID, Comm. I INFN, Cagliari, 13 settembre 2000 77
LNLFabric Design Detailed Program (2000-2001)Fabric Design Detailed Program (2000-2001)
Fabric ArchitectureFabric Architecture– Network topologiesNetwork topologies– Data Server connections and network file systemsData Server connections and network file systems– System break downSystem break down
Communication protocols for high speed network fabricsCommunication protocols for high speed network fabrics
Storage SystemsStorage Systems– Ultra SCSI (160/320/…)Ultra SCSI (160/320/…)– Ultra and Serial ATAUltra and Serial ATA– Storage Area Network (SAN)Storage Area Network (SAN)
Requirements document and survey of existing tools and Requirements document and survey of existing tools and technologies (month 6)technologies (month 6)
A configuration and installation management demonstrated to A configuration and installation management demonstrated to work on a cluster of more than 100 nodes (month 12)work on a cluster of more than 100 nodes (month 12)
A fully deployed service level monitoring system for a computer A fully deployed service level monitoring system for a computer centre. Hooks to provide remote requests for meta information centre. Hooks to provide remote requests for meta information like policies and quality measures, to allow schedule decisions like policies and quality measures, to allow schedule decisions (month 24)(month 24)
A fully integrated system to accept remote resource requests in A fully integrated system to accept remote resource requests in the form of tape mounts r jobs to run and provide monitoring the form of tape mounts r jobs to run and provide monitoring information about progress og requests, and final accounting information about progress og requests, and final accounting report back to the sender (month 36)report back to the sender (month 36)
G. Maron, INFN-GRID, Comm. I INFN, Cagliari, 13 settembre 2000G. Maron, INFN-GRID, Comm. I INFN, Cagliari, 13 settembre 2000 1212
LNLNFS Disk Servers Based Farm (Legnaro)NFS Disk Servers Based Farm (Legnaro)
Disk Server:Disk Server:-Dual PIII 800 MHzDual PIII 800 MHz- Mem 512 MBMem 512 MB- SCSI AdaptersSCSI Adapters
LNL is testing this farm module (with PIII at 450 MHz)LNL is testing this farm module (with PIII at 450 MHz)
DISKDISKSERVERSERVER
This Farm has been fundedThis Farm has been fundedBy Comm Calcolo for LNLBy Comm Calcolo for LNLOff-line computation (gr. 2/3)Off-line computation (gr. 2/3)
Requests for 2000:Requests for 2000:- 2 Raid SCSI Controllers - 2 Raid SCSI Controllers 10 Ml10 Ml
ANISANISBootBoot
ServerServer
G. Maron, INFN-GRID, Comm. I INFN, Cagliari, 13 settembre 2000G. Maron, INFN-GRID, Comm. I INFN, Cagliari, 13 settembre 2000 1313
LNLLNL PC Farm (phase I – May 200)LNL PC Farm (phase I – May 200)
Full-duplex 2.5+2.5 Gigabit/second links, switch ports, and Full-duplex 2.5+2.5 Gigabit/second links, switch ports, and interface portsinterface ports
Flow control, error control, and “heartbeat” continuity Flow control, error control, and “heartbeat” continuity monitoring on every linkmonitoring on every link
G. Maron, INFN-GRID, Comm. I INFN, Cagliari, 13 settembre 2000G. Maron, INFN-GRID, Comm. I INFN, Cagliari, 13 settembre 2000 2626
LNLInfiniBand (II)InfiniBand (II)
Legacy host architectureLegacy host architecture
The Infiniband ModelThe Infiniband Model
G. Maron, INFN-GRID, Comm. I INFN, Cagliari, 13 settembre 2000G. Maron, INFN-GRID, Comm. I INFN, Cagliari, 13 settembre 2000 2727
LNLInfiniBand (III)InfiniBand (III)
What?What?
– Initial single link signaling rate of 2.5Gbaud Initial single link signaling rate of 2.5Gbaud
Means unidirectional transfer rate of 250MB/sec with a Means unidirectional transfer rate of 250MB/sec with a theoretical full duplex rate of 500MB/sectheoretical full duplex rate of 500MB/sec
– Initial support for single, 4, and 12 wide link widthsInitial support for single, 4, and 12 wide link widths
– Point to point switched fabricPoint to point switched fabric
– Message based with multicasting supportMessage based with multicasting support
Simple test system (4 servers + Storafe Area Network + Simple test system (4 servers + Storafe Area Network + Network) for Network) for 20012001 is possible is possible
Early access to the productsEarly access to the products
Test Beds Test Beds
Requests 2001 (valuations) Requests 2001 (valuations) Comm V (Sadirc2000)Comm V (Sadirc2000)• 4 servers4 servers 50 Ml50 Ml• 1 IBA Switch1 IBA Switch 20 Ml20 Ml• IBA AdaptersIBA Adapters 10 Ml10 Ml
G. Maron, INFN-GRID, Comm. I INFN, Cagliari, 13 settembre 2000G. Maron, INFN-GRID, Comm. I INFN, Cagliari, 13 settembre 2000 2929