BeSTGRID Auckland Cluster
From BeSTGRID
The new BeSTGRID SGI Altix XE based Auckland Cluster has been ordered (23/01/2008) and expected arrival is late February 2008.
This SGI Altix XE [1] cluster is based on:
- Intel® Quad-Core Xeon® Processor-based architecture
- 1600 MHz front-side bus,
- 16GB of memory per compute node,
- an ultra-dense architecture that packs sixteen cores in a slim 1U form factor,
- fully-buffered DDR2 memory,
- driving the cluster with an Altix XE250 head node for advanced extensibility, redundancy, reliability, and I/O rich-features.
Contents |
[edit] Configuration
[edit] Hardware
| SGI Altix XE250 | SGI Altix XE320 | |
| Purpose | Head Node | Compute Nodes |
| Quantity | 1 | 5 systems x 2 Nodes each = 10 Nodes |
| Processors | 2x Dual Core Xeon 3.4GHz 6M 1600 FSB 80W X5272 (Harpertown) | 2x Quad Core Xeon 2.8GHz 12M 1600 FSB 80W E5462 (Harpertown) = 8 Cores per Node |
| Total Cores | 4 | 80 |
| Memory | 16GB Fully Buffered DIMM 800 MHz | 16GB Fully Buffered DIMM 800 MHz for each Node, 2GB per Core
= 160 GB memory |
| Disk space | 4x 750GB SATA with HW RAID | 1x 250GB SATA for each Node |
See here for more detailed configuration
[edit] Operating System
Linux CentOS is a standard de-facto for BeSTGRID computing environment. CentOS 5 is installed on the production cluster.
[edit] Cluster Software
The following Cluster software is installed:
- Rocks 5 Cluster Suit
- Specific functional packages are aprt of Rocks:
[edit] Environment
The following area available on the cluster:
- gcc
- subversion
- jdk 1.5. Created semilinks for java and javac in /usr/bin
- readline-devel and libtercap-devel (needed by mrBayes)
- mrBayes (without mpi, without Cantebury fixes)
- gcc4-c++.i386 rpm with yum (not from roll repos since couldn't satisfy dependencies.)
- gtk2-devel.i386 (not from rolls)
- lamarc (with default wxWidgets)
- emacs
- openmpi (not from rolls, but all rpms were there)
- cluslaw
- cluslaw mpi version
- modelTest
- cvs
- rpm-build
- VDT client
[edit] Computation Packages
The following packages are installed and maintained on the cluster:
| RPM | Software | Architecture | Patches | Dependencies | Build Process | Install Process | Location |
|---|---|---|---|---|---|---|---|
| beast-1.4.7-1 | BEAST | 32 bit | None | java 1.5 | none | Create symlinks for bin | Devbox, Production cluster |
| mrbayes-3.1.2-1 | MrBayes | 32 bit | None | readline-devel and libtercap-devel | make | Create symlink for mb | Devbox, production cluster |
| cluslaw-1.8-1 | cluslaw | 32 bit | None | none | make | Create symlinks | Devbox, Production cluster |
| cluslawmpi-0.13-1 | cluslaw (MPI version) | 32 bit | None | openmpi | make | Create symlinks | Devbox, Production cluster |
| lamarc-2.1.2b-1 | Lamarc | 32 bit | None | gtk2-devel | configure --enable-buildwx --prefix=location; make | make install | Devbox, Production cluster |
| modeltest-3.7-1 | ModelTest | 32 bit | None | None | make (need tarball) | create symlinks | Devbox, Production cluster |
[edit] Installation
[edit] Naming
The cluster is known as the BeSTGRID Auckland Cluster, with the canonical name being:
- hpc-bestgrid.auckland.ac.nz
There is also a test cluster. It's been build from three virtual machines in Xen Enterprise virtual environment with the canonical name being:
- vhpc-bestgrid.auckland.ac.nz
[edit] Hosting Location
The Cluster will be hosted within the Tier 1 University of Auckland ITS Data Centre, within the Owen G Glenn Building.
[edit] Testing
In order to confirm the Cluster is operating correctly to meet the Grid Operating Centre (GOC) requirements, testing has been carried out, documented in Auckland Cluster Testing
[edit] Operations
The BeSTGRID Auckland Cluster is monitored within the Grid Australia Grid Operations Centre (GOC), and activity is viewable via Ganglia.
[edit] Projects
[edit] Ocean Biogeographic Information System (OBIS)
- Project Lead
- Dr Mark J. Costello
- Usage
- OBIS virtual machine on Pleyads, 20GB of disk space.
- Project Description
- OBIS publishes data on behalf of scientists from government agencies, museums, universities, commercial companies, and non-governmental organisations. OBIS is always seeking new contributors.
- OBIS is a marine biogeographic information system, meaning that we concentrate on datasets that record particular species (or higher taxonomic group) from particular marine locations, at particular times. At present, we can only publish data where the locations are recorded as latitude and longitude, not as place names. Our focus is on high taxonomic quality, so datasets where organisms have been identified by professional or trained biologists are our priority. In the near future, we will be expanding to take in environmental datasets (i.e. coverage of physical, chemical, and geological parameters) that are relevant to understanding the distribution of species. So we are interested in hearing from potential contributors of these datasets, and welcome your contact, but are still in the process building this facility.
[edit] Austronesian Basic Vocabulary and Bantu Language Databases
- Project Lead
- Prof Russell Gray
- Usage
- .5 TB
- Project Description
- Austronesian Basic Vocabulary Database: This database contains 125,000 lexical items from 580 languages spoken throughout the Pacific region. These languages all belong to the Austronesian language family, which is the largest family in the world. It contains between 1,000 and 1,200 languages.
- Bantu Language Database: This database contains 2,388 lexical items from 6 Bantu languages
