Personal tools


Computational Resources
The Advanced Computing Laboratory is the flagship of computational resources for TLC2 . Below are listed a few featured systems for this high performance computing lab.

To apply for an account, fill in the application (pdf) and return the completed form to room 218 in PGH.
To see our current running jobs and your place in the queue,

Additional cluster related documentation, tutorial, and handout are available here.


TLC2 Itanium2 Cluster
RESOURCE

The HP Itanium2 based Atlantis cluster is owned by the Texas Learning and Computation Center (TLC2). The cluster currently runs RedHat Enterprise Linux 3 for the ia64 architecture.

To see our current running jobs and your place in the queue for the Itanium2 Clusters go to: http://sierra.tlc2.uh.edu/ganglia/addons/job_monarch/?c=Atlantis.

HARDWARE

Node Type
CPU Type
CPU Count
Memory
Node Count
login
Itanium2 (1.4GHz)
1
4GB
2
dual
Itanium2 (1.3GHz)
2
4GB
152
quad
Itanium2 (1.3GHz)
4
32GB
48
16-way SMP
Itanium2 (1.5GHz)
16
32GB
1


INTERCONNECTS

There are three interconnects present: myrinet, gigabit Ethernet and fast Ethernet. The myrinet network is the primary high performance cluster interconnect. The gigabit Ethernet network is intended for storage traffic, some management, and connectivity outside the cluster with Globus. The fast Ethernet is strictly NIS, multicast traffic and management and is not available for job runs.

Usage guide (PDF)

TLC2 SGI Altix
TLC2 operates two SGI Altix 3700's. These machines are currently running SUSE Linux Enterprise Server 10 with SGI ProPack 5 for the ia64 architecture. Both machines have 64 1.3GHz Itanium2 processors and 512GB RAM running as a single system image.

To see the current status of the job queue go to: http://flexo.tlc2.uh.edu/ganglia/addons/job_monarch/?c=Flexo.



Usage guide (PDF)

Eldorado Itanium2 Cluster
The Advanced Computing Research Laboratory has implemented a 61 node Itanium2 cluster with a peak capacity of 472 Gflops and 256 GB of total memory. The cluster interconnect is Clusteredge from Scali.  The cluster configuration is as follows: 56 dual processor nodes with 900 MHz processors, 2 nodes with 1.3 GHz processors, and 2 nodes with 1.5 GHz Itanium2 processors.
  Each dual processor node is configured with 4 GB of memory and has a system bus bandwidth of 6.4 GB/sec and a 8.5 MB/sec memory bandwidth. The nodes have a 36 GB 10,000 rpm SCSI disk and three 64-bit 133 MHz PCI buses.  The 900 MHz nodes have a peak capacity of 7.2 Gflops, the 1.3 GHz nodes 10.4 GFlops, and the 1.5 GHz nodes 12 GFlops. The quad processor node has four 1.5 GHz Itanium2 processors and 16 GB of memory, with a system bus bandwidth of 6.4 GB/sec and a memory bandwidth of 12.8 GB/sec.  This node has two 36 GB 10,000 rpm SCSI disks and three 64-bit 133 MHz PCI buses and a peak capacity of 16 Gflops. The Clusteredge interconnect has been demonstrated to achieve a bandwidth of up to 382 MB/sec per node with a latency of about 4 microseconds at the application level (MPI).

The cluster is currently running CentOS 4.4.

Usage guide (PDF).



ACRL/TLC2 Opteron Cluster
The Advanced Computing Research Laboratory and TLC2 have purchased an AMD Opteron cluster from Microway. The current cluster consists of 20 nodes:

- 16 dual AMD Opteron(tm) Processor 246 systems, each with 4GB of ram

- 4 quad AMD Opteron(tm) Processor 842 systems, each with 16GB of ram

The cluster nodes are connected with both an Infiniband and gigabit interconnect.

The cluster is running Rocks 4.2.1.

To see our current running jobs and your place in the queue for the Opteron Medusa Cluster go to: http://medusa.tlc2.uh.edu/ganglia/addons/rocks/queue.php

Usage Guide (pdf)

Storage resources

There are 22.8 Tera Bytes of raw storage available to our clusters, including 7.6 TB of SCSI RAID storage and 15.2 TB of Lustre/NFS storage. In addition, Distributed Storage Network project provides 37.7 TB of raw storage to several research centers on campus.


Networking

TLC2 has established High-Performance networks serving both its investigators and campus at large. TCCIS, the center that preceded TLC2, through its director successfully acquired NSF funding in 1996 for connecting UH to the NSF national high-speed backbone, the vBNS, and established one of the first national gateways to high-performance national and international computer networks, the Texas GigaPoP jointly with Rice University and Baylor College of Medicine. The Texas GigaPoP has since evolved into the South-East Texas GigaPoP (SETG) with additional members.

The second High-Performance network established by TLC2 was the Distributed Storage Network mentioned above.

The third High-Performance network established by TLC2 is the Research and Education Network of Houston, RENoH, based on dark fiber connecting UH, Rice, and the Texas Medical Center with each other, as well as the state network LEARN and the national high-performance backbone networks, the National Lambda Rail (NLR) and Internet2. RENoH currently has about 500 miles of fiber covering about 22 miles.

Plans are underway to extend this network to the UH Clear Lake campus, NASA, UTMB Galveston and Texas A&M Galveston. This expansion of RENoH is expected to be completed by the end of 2007.

TLC2 has acquired its own AS (Autonomous System) number and its own internet protocol (IP) address range acquired from IANA (Internet Assigned Numbers Authority) enabling routing and security policies for research depending on computer networks to be established independent of the needs of campus at large. TLC2 computational, storage and visualization resources are connected to the TLC2 network as are center desktop computers, printers and servers.