ncar s globally accessible data environment (glade) · ncar’s globally accessible data...
TRANSCRIPT
NCAR’s Globally Accessible Data
Environment (GLADE)
Globus World 2014 16 April 2014
Pamela Gillman, NCAR Manager, Data Analysis Services Group
Globus World 2014 16 April 2014
Data Analysis Services Group
l Data Transfer and Storage Services l Pamela Gillman l Joey Mendoza l Craig Ruff
l High-Performance File Systems
l Data Transfer Protocols
l Visualization Services l John Clyne l Alan Norton l Scott Pearse l Miles Rufat-Latre (student)
l VAPOR development and support
l 3D visualization
NCAR / CISL / HSS / DASG
Globus World 2014 16 April 2014
GLADE GLobally Accessible Data Environment l Unified and consistent data environment for
NCAR HPC l Supercomputers, Data Analysis and Visualization Clusters l Support for project work spaces l Support for shared data transfer interfaces l Support for Science Gateways and access to ESG & RDA data
sets
l Data is available at high bandwidth to any server or supercomputer within the GLADE environment
l Resources outside the environment can manipulate data using common interfaces
l Choice of interfaces supports current projects; platform is flexible to support future projects
Globus World 2014 16 April 2014
GLADE Environment
GLADE
HPSS
Data Transfer Gateways Science Gateways
Project Spaces Data Collec;ons $HOME $WORK
$SCRATCH
Remote Visualiza;on
Computa;on Analysis & Visualiza;on
RDA ESG CDP
geyser caldera pronghorn
yellowstone
Globus Online GridFTP
HSI / HTAR scp, sOp, bbcp
VirtualGL
Data Share
Globus World 2014 16 April 2014
GLADE Storage Overview l 10.5 PB useable
l + 6 PB useable, total 16.4 PB usable (April 2014)
l > 90 GB/s sustained bandwidth l 76 DCS3700 systems with 1 expansion chassis l 6840 3TB drives l SAS direct connect to NSD servers l 20 NSD servers, 6 management nodes l 2 InfiniBand management nodes l 4 data transfer nodes l 1 108-port IB FDR 14 switch, 6 ethernet switches l 21 racks l GPFS Parallel File System
Globus World 2014 16 April 2014
DataShare Storage Overview
l 1.5 PB useable l > 5 GB/s sustained bandwidth l 1 DDN 9900 system l 300 1TB drives, 900 2TB drives l 4 NSD and management nodes l DDR IB direct connect from storage to servers l 3 racks l GPFS Parallel File System l Globus Plus integration
Globus World 2014 16 April 2014
Data Service I/O Networks
Remote Visualization VirtualGL TurboVNC
HPSS ESG Science Gateway
CDP Science Gateway
FDR InfiniBand
10Gb Ethernet
Yellowstone
ys0101 ys0201
ys0301 ys0401
ys0401 ys0401
yslogin yslogin
yslogin
4536 nodes
GLADE 6Gb SAS
nsd1 nsd20
16.4 PB
geyser DAV Cluster
caldera 32 nodes
OpenGL GPGPU
> 90 GB/s 110 GB/s
pronghorn
Data Transfer Services
gladedm
4 nodes
Globus, GridFTP scp, sftp, bbcp
HSI/HTAR
10 GB/s
gladedm gladedm
gladedm
DATASHARE
DDR IB nsd1 nsd4
1.5 PB
RDA Science Gateway
“Phi”
Globus World 2014 16 April 2014
GLADE Data Workflow Solutions
l Information centric data model l Data can stay in place through entire workflow l Access from supercomputing, data post-
processing, analysis and visualization resources l Direct access to NCAR data collections
l Availability of persistent longer-term storage l Allows completion of entire workflow prior to final
storage of results either at NCAR or offsite l Provides high-bandwidth data transfer and
data sharing services between NCAR and peer institutions
Globus World 2014 16 April 2014
GLADE Growth
0
1,000
2,000
3,000
4,000
5,000
6,000
7,000
TB
/glade/p/work /glade/project /glade/scratch
As of March 29, 2014: • /glade/p/work holds 160 TB • /glade/scratch holds 3,085 TB • /glade/project holds 2,850 TB
Globus World 2014 16 April 2014
CISL Science Gateway Support
l Research Data Archive l 1 PB allocation l sub setting services are performed on geyser l direct access to online data collections from batch
jobs l Earth Systems Grid / Community Data Portal
l 1 PB allocation l direct access to CMIP5 and NARCAP data from
batch jobs
Globus World 2014 16 April 2014
Data Transfer Gateway
HPSS
GLADE
Data Share
gladedm1 gladedm2 gladedm3 gladedm4
data-access.ucar.edu 10GB/s network bandwidth
FDR IB
10 Gb
•
LSF job
queue for
HSI /
HTAR access to
HPSS
•
GridFTP /
Globus
Online endpoints for large transfers
•
scp/sftp,
>5GB/s
>90GB/s
nsd1 nsd2 nsd3 nsd4
Globus World 2014 16 April 2014
Data Transfer Services
l Globus Online Endpoints l launch and forget data transfers l Access with users UCAS account and token
l ncar#gridftp l Access with users XSEDE Account
l xsede#ncar l xsede#glade
l Web UI, CLI, REST API l Globus Connect for transfer to/from your desktop
l gridftp, globus-url copy, scp/sft, bbcp l HSI/HTAR for HPSS access through LSF
Globus World 2014 16 April 2014
Data Sharing Services
l ncar#datashare l Globus Plus implementation l data sharing allocations for self-publishing or
data delivery l data owner controls access
l can create groups for access control l can share ‘read-only’ or ‘read-write’
l user can create custom access interfaces l CLI or REST API
Globus World 2014 16 April 2014
Data Sharing Use Cases
l Delivery of data from non NCAR users for publication in a Science Gateway
l Delivery of 3D visualization to non NCAR users
l Publication of supporting data associated with publication
l Share a file or data set with a non NCAR collaborator
Globus World 2014 16 April 2014
Data Sharing Futures
l User Outreach and Education l need more publicity for available services l schedule a user training seminar
l Potential to expanding ‘Sharing’ capability to the larger project spaces
l Potential to couple GlobusOnline more tightly with the Science Gateways
l Potential project to help build a custom UI for access to a data collection
l Re-evaluation HPSS integration with Globus
Globus World 2014 16 April 2014
QUESTIONS? [email protected]