Computing Resources
  1. Task A
    1. ATLAS
    2. D0
  2. Task D
    1. Mantrid - used for E852, Radphi analysis and GlueX/Hall D Monte Carlo
    2. AVIDD - This is a 200-processor cluster at the IU Computing Center which is being used by the Task D group for data analysis. This facility was recently dedicated and the the Task D use for data analysis was showcased.
  3. Lattice QCD

D0

For D0, in addition to desk-top PC's, an Institutional Analysis Center (IAC) has been set up at Indiana. The center consistes of a front-end node and five dual-processor Linux machines. It also includes a RAID disk server and disks with total storage capacity to 1.4 TB. This center provides a remote D0 software development environment, remote analysis of data, remote reconstruction and production, and Monte Carlo generation and reconstruction. This IAC is a ``station" of the D0 Sequential data Access via Meta-Data (SAM) system that allows for efficient and transparent data delivery and sharing, alleviating the load on central data storage and servers: the collaboration can seamlessly access data stored at Indiana and our compute power can be easily applied to the large data sets physically at FNAL. Half of two 800~GB RAID arrays are available exclusively to the group members resident at FNAL.

ATLAS

Through an in-kind contribution from Indiana University, in 2001the iVDGL project received hardware for an initial configuration for a prototype ATLAS Tier 2 center located at IU. This center was initially located in Indianapolis where high speed access to the Internet2 network (Abilene) was available on the Indiana University - Purdue University Indianapolis (IUPUI) Campus. When a dedicated optical fiber connection between Indianapolis and Bloomington became available in 2002, the Tier 2 center was relocated to the Wrubel Computing Center here on the IU Bloomington Campus. The center consists of a dual CPU head node and sixteen dual CPU worker nodes. The CPUs are 400 MHz Pentium IIs and each node has 1 GB of memory shared between the 2 CPUs except for the head node which has 2 GB of memory. The head node and worker nodes are connected to an HP 4000 fast ethernet switch to communicate with each other. In addition there is a 1.8 GHz P4 machine used to providing an AFS server and aa Unix workstation that serves 200 GB of disk space to the cluster via NSF. The AFS server and head node network connections are 1 Gb/s to the HP switch and the outside world. All other connections are 100 Mb/s. An additional in-kind contribution of 0.5 FTE from the University has provided the effort used to setup the Tier 2 center, install standard grid middleware tools such as globus and condeor, and do routine maintenance. This initial cluster configuration has been used extensively for ATLAS data challenge Monte Carlo production and development of the grid portal Grappa.

In the first half of 2003 this initial configuration of the prototype IU Tier 2 center will be retired and the IU Tier 2 center will relocate on the AVIDD research computing cluster that IU commissioned in March of 2003 as an University wide computer cluster). AVIDD is funded by a $1.8M MRI grant from NSF and consists 96 compute nodes in Bloomington and 96 compute nodes in Indianapolis. The compute nodes have dual CPU 2.4 GHz Xeon CPUs, 512 kB cache and 2 GB RAM. In addition there are 2 head nodes and 4 storage nodes. The storage is fast SCSI disks connected to optical fiber. Within AVIDD, Myrinet is used internally move data between the disks and CPUs. Equipment money from the NSF-funded iVDGL grant will be used to buy an additional 6 cabinets of fast disks in exchange for preemptive scheduling for the Tier 2 center on 32 nodes (64 CPUs). These new disks are 10,000 rpm SCSI disks, contain a total 1.5 TB, and are connected to the cluster using a very fast fiber channel channel interface. The new disks will be divided into a 1 TB partition used for scratch purposes by ATLAS/iVDGL and a 0.5 TB partition for user home directories. It will be possible for ATLAS/iVDGL to opportunistically use any unused CPUs in both the Indianapolis and Bloomington clusters. It will also be able to request a bigger allocation of computers during intense running periods. All of the nodes are connected to Abilene via 1 Gb/s connections.

Mantrid is a cluster of 18 computers used by Task D for data analysis. Each computer has:

  • 2 Intel P3-800MHz CPUs
  • 256 MB RAM
  • 90 GB hard drive space
  • 100base-T ethernet
  • RedHat 8.0 Linux operating system

Lattice QCD Calculations

Cluster used by Steve Gottlieb