When possible, we would greatly appreciate an acknowledgement of ACCRE in presentations and publications. Here is our suggested text (also available as an RTF document):
We also attempt to maintain on this website an up-do-date list of publications based on your research using ACCRE that results in journal articles or conference proceedings.
We are glad to assist you with the development of proposals that include the use of ACCRE resources. We will also provide a letter in support of the portion of your research and educational programs which use the facility. Please do not hesitate to contact ACCRE Administration for any such assistance, including budgetary concerns.
For your convenience, a general description of ACCRE facilities is available below and as an RTF document. Please feel free to use any or all of the included text. For more technical detail, see the services pages and this FAQ.
The Advanced Computing Center for Research and Education (ACCRE) is built and operated by Vanderbilt faculty. ACCRE offers computing resources flexible enough to enable High Performance Computing applications in a wide variety of research and education areas. All ACCRE hardware resources are housed in the University's secure data center and administered by a team of ACCRE system administrators. In addition to the High Performance Computing system (described in detail below), ACCRE has mulitple terabytes of disk space and a robotic tape storage system. Vanderbilt University is an Internet2 member and a participant in the Abilene network.
The ACCRE High Performance Computing system consists of 776 x86 processors (160 2.4 GHz Opteron processors, 456 2.0 GHz Opteron processors plus 160 1.8 GHz Opteron processors) and 644 PowerPC processors (2.2 GHz IBM JS20 Blades), both running a 64-bit Linux OS. Each processor has at least 1 GB of memory, a 40 GB disk drive, and dual copper gigabit Ethernet ports. Over one-third of the systems also have Myrinet networking. Each node is monitored via Nagios. The cluster has over 1400 processors and the theoretical peak performance is roughly 7 TFLOPS. The gateways are equipped with external gigabit Ethernet connections. Resource management, scheduling of jobs, and tracking usage is handled by the Moab/Torque schedulers. These utilities include an "advance reservation" system that allows a block of nodes to be reserved for pre-specified periods of time (e.g., a class or lab session) for educational purposes.
IBM's General Parallel File System (GPFS) is used for user home directories and scratch space and can sustain 10 Gb/s to the cluster. The home directories of all users are backed-up daily. The internal network design is a classic tree with a Foundry BigIron RX-16 switch at the root providing 1.54 Tb/s of available bandwidth (3.84 Tb/s total switch capacity) to the cluster. All of the disk servers and management nodes are connected to the top level. A collection of compute nodes, typically 20, are connected to a small local gigabit switch with a 1 Gb uplink to the RX-16.
ACCRE's suite of computing resources also has one high-memory 16-processor (1.5 GHz Itanium2) SGI Altix with 6MB cache per processor and a total of 32 GB memory.
