The Scalable Computing Software (SCS) Laboratory in the Computer Science Department at Illinois Institute of Technology are funded by a DoD DURIP grant (DAAD19-01-1-0432), three NSF CISE-RR grants (EIA-0130673, EIA-0224377, CNS-0751200), plus industry, state, and institutional matching for its computing infrastructure. Its computing facility includes an 64-node Sun Microsystems ComputeFarm, a 17-node Dell cluster, a 14-node IBM Linux-based cluster, a 12-node Cray XD1 supercomputer, a 20-node Opteron cluster (Donated from MicroSoft), a 72-processor SiCortex cluster, a large Access Grid node, several multi-processor file servers, many high-end graphic workstations, and other advanced computing and communication facilities.
The Sun ComputeFarm is composed of one Sun Fire X4240 server, one Sun Enterprise 450 server, 64 Sun Fire X2200 and 20 Sun Fire V210R compute nodes, with a totally of 172 processors and 562 cores. The nodes are connected fully with Gigabits Ethernet and partially with InfiniBand. The model of head node is Sun Fire X4240, with dual 2.7 GHz Opteron quad-core processors, 8GB memory, and 12 500GB 7200RPM SATA-II drives configured as RAID5 disk array. The model of computing nodes is Sun Fire X2200 servers, each node with dual 2.3GHz Opteron quad-core processors, 8GB memory, and a 250GB 7200RPM SATA hard drive. We added a SSD on each node during this March and April 2011. The SSD model is: OCZ Technology OCZSSDPX-1RVDX0100 REVO X2 PCIE SSD 100GB MLC. In April, we also equipped two nodes with Electronic Educational Devices 72222 Watts Up? .net electricity meter with wired/wireless remote Internet monitoring.
The Sun Fire cluster are used as experimental platform for research on parallel I/O, including data access cost model analysis, data layout optimization in parallel file systems, coordinated I/O scheduling in multiple servers, cache and prefetching for high-performance and data-intensive computing, fault-tolerance in HPC and Cloud Computing, storage systems for cloud computing, simulation on memory and cache architectures, etc.. They are also used as testbed for students course projects from CS department.
REVO X2 PCIE SSD
Electronic Educational Devices 72222 Watts Up?
The Dell cluster is composed of one Dell PowerEdge 2850 head node, 16 Dell PowerEdge 1425 nodes. All of the Dell PowerEdge 1425 Nodes are connected with a Gigabist Ethernet.
The IBM cluster is composed of 14 IBM x335 xseries servers with a totally 28 processors. It is connected to and shares advanced computing facilities at the National Center for Supercomputing Applications (NCSA), the Argonne National Laboratory, Northwestern University, the University of Chicago, and the University of Illinois at Chicago via a 7.5-million-dollar dedicated 10-gigabit-per-second fiber optic research data network to form the Distributed Optical Testbed (DOT). DOT is a distributed teraflops/gigabit testbed. It provides an excellent computing backbone for the proposed research.
The Opteron Cluster is composed of 20 Opteron 270 server nodes. Each node is equipped with two AMD Opteron 246 CPUs, 8G memory and 1TB disk. All the nodes are connected via Gigabit Ethernet.
SiCortex SC072-PDS is a complete 72-processor cluster packaged in a whisper-quiet, low-powered deskside cabinet using less than 300 watts of power. Each SC072-PDS has a 48 GByte memory system, two gigabit Ethernet ports, and up to three (3) optional PCI Expressâ„?cards, each operating at a peak speed of 2 GBytes/sec. Each of the twelve (12) SC072-PDS nodes is a multi-core chip that contains six (6) processor cores, each operating at 1.4 GFlops, together with all the switch fabric circuits necessary to support interprocessor communications. The nodes are connected in the Kautz topology that is the hallmark of SiCortex systems.
In the last few years, our bigger machines become even bigger and small machines become much smaller.
Bigger becomes even bigger:
 
 
Smaller becomes much smaller:
 
And somethings never change:
|
