Kurt Mueller

1
NPACI HotPage Updates

• Kurt Mueller
• San Diego Supercomputer Center
• kurt_at_sdsc.edu

2
NPACI
National Partnership for Advanced Computational
Infrastructure

• US National Science Foundation funded
  collaboration between research institutions
• San Diego Supercomputer Center (SDSC) -
  University of California, San Diego
• Texas Advanced Computing Center (TACC) -
  University of Texas, Austin
• Center for Advanced Computing (CAC) - University
  of Michigan
• 46 other institutions...

3
NPACI resources

• compute
• SDSC Blue Horizon - 1152 CPU IBM SP
• TACC Longhorn - 224 CPU IBM Power4
• TACC Aurora - 16 CPU Cray SV1
• CAC Morpheus, Hypnos - 256, 134 CPU AMD clusters
• CAC 64 and 24 CPU IBM SP systems

4
NPACI resources

• storage
• SDSC HPSS - 6 petabytes
• TACC DMF - 1.4 petabytes
• CAC ADSM - 41 terabytes
• Caltech HPSS - 323 terabytes

5
aggregate resources

• 463 nodes
• 1854 CPUs
• 4345 peak GFlops
• 1714 GB RAM
• 7.7 petabytes tape storage

6
(No Transcript)
7
status bar
8
individual resource info
9
login for more tools
10
architecture
11
HotPage updates

• moved to Apache on linux
• underlying GridPort updated to 2.3
• object-oriented perl
• module packaging
• simplified installation
• new features NWS, etc.

12
data collection

• old methods of data collection
• cron jobs on webserver, triggering ssh commands
  to execute remote scripts on resources and store
  data on webserver
• Globus job run to execute remote scripts
• GIS queries for status data

13
data collection

• problems with old methods
• some scripts generated LDIF, some HTML
• no separation of content and presentation
• GIS support inconsistent across partner sites
• no persistent storage of data - historical record

14
GP-IR data repository

• GridPort Information Repository
• developed at TACC, with help from SDSC
• XML-centric database to store performance and
  other data
• web services data ingester and query interface

15
(No Transcript)
16
GP-IR data example

• lt?xml version"1.0" encoding"UTF-8" ?gt
  ltloadInfo xmlns"http//grids.tacc.utexas.edu/
  schemas/infoservices/load" xmlnsxsi"http//www.w
  3.org/2001/XMLSchema-instance" xsischemaLocation
  "http//grids.tacc.utexas.edu/schemas/infoservices
  /load/load.xsd" timestamp"2001-12-17T093047-05
  00" hostname"horizon.sdsc.edu"gt
• ltloadgt
• lttypegtqueuelt/typegt
  ltPercentgt87lt/Percentgt
• lt/loadgt
• lt/loadInfogt

17
GP-IR data caching

• cron jobs run on HotPage server to query GP-IR
  every five minutes and write out xml to cache
  directories
• HotPage page requests generate live queries to
  GP-IR
• if live queries fail, HotPage scripts load cached
  XML

18
resource view sorting
19
grid view

• summary table
• grid monitor
• NWS

20
to do

• historical performance queries
• portlets - presentation this week
• customization, portability
• GridPort 3.0 - ditto
• OGSA
• HotPage source code distribution - this week?
• rollout technologies to PACI HotPage
• NPACI, NCSA and PSC resources

21
links

• https//hotpage.npaci.edu/
• http//www.npaci.edu
• http//www.sdsc.edu
• http//www.tacc.utexas.edu
• GridPort http//gridport.npaci.edu
• GP-IR http//www.tacc.utexas.edu/grid/gpir/