1 scientific data management (sdm) center for enabling technologies (cet) lead institution: lbnl...
TRANSCRIPT
1
Scientific Data Management (SDM)Scientific Data Management (SDM)
Center For Enabling Technologies (CET)Center For Enabling Technologies (CET)
Lead Institution: LBNLLead Institution: LBNL
Coordinating PI: Arie ShoshaniCoordinating PI: Arie Shoshani
2
Scientific Data Management CenterScientific Data Management Center
Center PI: Arie Shoshani LBNL
DOE Laboratories co-PIs:
Bill Gropp, Rob Ross* ANLArie Shoshani, Doron Rotem LBNLTerence Critchlow*, Chandrika Kamath LLNLNagiza Samatova* ORNL
Universities co-PIs :Mladen Vouk North Carolina State Alok Choudhary Northwestern Bertram Ludaescher, Ilkay Altinas UC Davis + SDSCSteve Parker U of Utah
* Area Leaders
Participating Institutions
3
A Typical SDM ScenarioA Typical SDM Scenario
Control Flow Layer
Applications &Software Tools
Layer
I/O System Layer
Storage & NetworkResouces
Layer
Flo
w T
ier
Wo
rk T
ier
+
DataMover
SimulationProgram
ParallelR
PostProcessing
TerascaleBrowser
Task A:Generate
Time-Steps
Task B:Move TS
Task C:Analyze TS
Task D:Visualize TS
ParallelNetCDF
PVFS SabulHDF5
LibrariesSRM
4
Technology Details by LayerTechnology Details by Layer
Hardware, OS, and MSS (HPSS)
WorkFlowManagement
Tools
Web Wrapping
Tools
EfficientParallel
Visualization(pVTK)
Efficientindexing(Bitmap Index)
DataAnalysis
tools(PCA, ICA)
ASPECT:integration Framework
Parallel NetCDFSoftware
Layer
ParallelVirtual
FileSystem
StorageResourceManager
(To HPSS)
ROMIOMPI-IOSystem
DataMining &Analysis(DMA)Layer
StorageEfficientAccess(SEA)Layer
ScientificProcess
Automation(SPA)Layer
Hardware, OS, and MSS (HPSS)
WorkFlowManagement
Tools
Web Wrapping
Tools
EfficientParallel
Visualization(pVTK)
Efficientindexing(Bitmap Index)
DataAnalysis
tools(PCA, ICA)
ASPECT:integration Framework
Parallel NetCDFSoftware
Layer
ParallelVirtual
FileSystem
StorageResourceManager
(To HPSS)
ROMIOMPI-IOSystem
DataMining &Analysis(DMA)Layer
StorageEfficientAccess(SEA)Layer
ScientificProcess
Automation(SPA)Layer
Analysis
Parallel R
Statistical
5
Example Data Flow in TSIExample Data Flow in TSI
InputData
HighlyParallelCompute
Output~500x500files
Aggregate to ~500 files (< 2 to 10+ GB each)
Archive
Data Depot
Logistic NetworkL-Bone
Local MassStorage 14+TB)
Aggregate to one file (1+ TB each)
VizWall
Viz Client
Local 44 Proc.Data Cluster- data sits on local nodes for weeks
Viz Software
Logistical Network
Courtesy: John Blondin
6
Using the Scientific Workflow Tool (Kepler)Using the Scientific Workflow Tool (Kepler)Emphasizing Dataflow Emphasizing Dataflow (SDSC, NCSU, LLNL)(SDSC, NCSU, LLNL)
Automate data generation, transfer and visualization of a large-scale simulation at ORNL
7
FastBitFastBit
A compressed bitmap indexing A compressed bitmap indexing technology for efficient searching of technology for efficient searching of
read-only dataread-only data
http://sdm.lbl.gov/fastbithttp://sdm.lbl.gov/fastbit
8
FastBit OverviewFastBit Overview
• FastBit is designed to search multi-FastBit is designed to search multi-dimensional datadimensional data• Conceptually in table format
• rows objects• columns attributes
• FastBit uses vertical (column-FastBit uses vertical (column-oriented) organization for the dataoriented) organization for the data• Efficient for analysis of read-only data
• FastBit uses FastBit uses compressed bitmap compressed bitmap indicesindices to speed up searches to speed up searches• Proven in analysis to be optimal for single-
attribute queries• Superior to other optimal indices because they
are also efficient for multi-attribute queries
rowcolumn
9
Basic Bitmap IndexBasic Bitmap Index
• Compact: one bit per Compact: one bit per distinct value per objectdistinct value per object
• Easy to build: faster than Easy to build: faster than common B-treescommon B-trees
• Efficient to query: only Efficient to query: only bitwise logical operationsbitwise logical operations
• A < 2 b0 OR b1
• 2<A<5 b3 OR b4
• Efficient for multi-Efficient for multi-dimensional queriesdimensional queries• Use bitwise operations to
combine the partial results
Datavalues015312041
100000100
010010001
000001000
000100000
000000010
001000000
=0 =1 =2 =3 =4 =5
b0 b1 b2 b3 b4 b5
10
Grid Collector FeaturesGrid Collector Features
Key features of the Grid Collector:Key features of the Grid Collector:• Providing transparent object access• Selecting objects based on their attribute values• Improving analysis system’s throughput• Enabling interactive distributed data analysis
11
Grid Collector Speeds up Grid Collector Speeds up AnalysesAnalyses
0
1
2
3
4
5
0 0.2 0.4 0.6 0.8 1
selectivity
sp
ee
du
p
Sample 1
Sample 2
Sample 3
• Test machine: 2.8 GHz Xeon, 27 MB/s read speedTest machine: 2.8 GHz Xeon, 27 MB/s read speed
• When searching for rare events, say, selecting one event out of 1000, When searching for rare events, say, selecting one event out of 1000, using GC is using GC is 20 to 5020 to 50 times faster times faster
• Using GC to read 1/2 of events, speedup > 1.5, 1/10 events, Using GC to read 1/2 of events, speedup > 1.5, 1/10 events, speed up > 2.speed up > 2.
1
10
100
1000
0.00001 0.0001 0.001 0.01 0.1 1
selectivity
sp
ee
du
p
Sample 1
Sample 2
Sample 3
12
FastBit-Based Multi-Attribute Region FastBit-Based Multi-Attribute Region Finding is Theoretically OptimalFinding is Theoretically Optimal
On 3D data with over On 3D data with over 110 million points110 million points,,
region finding takes region finding takes less than 2 secondsless than 2 seconds
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
10000 110000 210000 310000 410000
Number of line segments
reg
ion
gro
win
g t
ime
(sec
)Flame Front discovery
(range conditions for multiple measures) in a combustion simulation (Sandia)
Time required to identify regions in 3D Supernova simulation (LBNL)
13
Objects On-Demand: Objects On-Demand:
from Files to Object Managementfrom Files to Object Management
A Scientific Application Partership (SAP)A Scientific Application Partership (SAP)
Lead Institution: BNLLead Institution: BNL
Coordinating PI: Jerome LauretCoordinating PI: Jerome Lauret
14
Participating InstitutionsParticipating Institutions
• Participating Institutions Participating Institutions • BNL : Jerome Lauret• LBNL : John Wu• SLAC: Andy Hanushevsky
• TechnologiesTechnologies• FastBit• SRM (DRM, HRM)• xrootd
15
Xrootd:Single Level SwitchXrootd:Single Level Switch
ClientClient RedirectorRedirector(Head Node)
Data ServersData Servers
open file X
AA
BB
CC
go to C
open file X
Who has file X?
I have
Cluster
Client sees all servers as xrootd data serversClient sees all servers as xrootd data servers
2nd open X
go to C
RedirectorsRedirectorsCache fileCache filelocationlocation
16
Xrootd:Single Level SwitchXrootd:Single Level Switch
ClientClient RedirectorRedirector(Head Node)
Data ServersData Servers
open file X
AA
BB
CC
go to C
open file X
Who has file X?
I have
Cluster
Client sees all servers as xrootd data serversClient sees all servers as xrootd data servers
2nd open X
go to C
RedirectorsRedirectorsCache fileCache filelocationlocation
DRM
DRM
DRM
HRM
MSS
archive
17
Objects on-demandObjects on-demand
xrootd
18
Storage Resource Management (SRM)Storage Resource Management (SRM)
Center For Enabling Technologies (CET)Center For Enabling Technologies (CET)
Lead Institution: LBNLLead Institution: LBNL
Coordinating PI: Alex SimCoordinating PI: Alex Sim
19
Participating InstitutionsParticipating Institutions
• BNL : Jerome LauretBNL : Jerome Lauret
• FNAL : Don Petravick, Timur PerelmutovFNAL : Don Petravick, Timur Perelmutov
• TJNAF : Andy KowalskiTJNAF : Andy Kowalski
• LBNL : Alex Sim, Arie ShoshaniLBNL : Alex Sim, Arie Shoshani
• UCSD : Abhishek Singh RanaUCSD : Abhishek Singh Rana
• U. of Wisc: Miron LivnyU. of Wisc: Miron Livny
20
Proposed workProposed work
• Development of new functional features as part Development of new functional features as part of the SRM collaboration (coordinated by LBNL)of the SRM collaboration (coordinated by LBNL)• Authorization• Monitoring• Performance estimation
• Development of new versions of SRMs by Development of new versions of SRMs by participating institutionsparticipating institutions• Disk systems and HPSS (LBNL)• dCache (FNAL)• Jasmine (TJNAF)
21
New AspectsNew Aspects
• Development of monitoring components for Development of monitoring components for bandwidth and networking availability bandwidth and networking availability (U. Wisc, FNAL)(U. Wisc, FNAL)• Better control of SRM behavior• Performance estimation
• Integration of Lambda station interface into the Integration of Lambda station interface into the SRM middleware (FNAL)SRM middleware (FNAL)
• Development of an authorization framework Development of an authorization framework (UCSD)(UCSD)• To enforce access privileges• used by SRMs for policy declaration by VO and Sites
22
SRM CollaborationSRM Collaboration
• Continued support of SRMs in experiments and projectsContinued support of SRMs in experiments and projects• ATLAS (BNL, FNAL)• CLAS (TJNAF)• CMS (FNAL)• CPES (LBNL)• ESG (LBNL)• Lattice QCD (TJNAF, FNAL)• Phenix (BNL)• STAR (BNL, LBNL)
• Coordination with other centers and institutesCoordination with other centers and institutes(including LCG, RAL, EGEE).(including LCG, RAL, EGEE).
• Goal: joint specification of SRM through regular Goal: joint specification of SRM through regular meetings, joint documents, and GGF participationmeetings, joint documents, and GGF participation