Sun Fire Cluster
|
The SunFire cluster is a symmetric multiprocessor (SMP) system based on the award-winning UltraSPARC-III processor and the robust Solaris 9 Operating Environment.
The initial four SunFire 6800 servers (SFNODE0 to SFNODE3) are configured with 24 x 1.05 GHz (8 MB E-Cache) UltraSPARC-III processors and 96 GB of memory. All four cluster nodes are running Solaris 8 (HW 02/02), Sun HPC ClusterTools 5.0,
Sun GridEngine 5.3 Enterprise Edition, and Forte Developer 6 (Update 2) and are connected using Gigabit ethernet. An additional four SunFire 6800 servers (SFNODE4 to SFNODE7) with 24 x 900 Mhz (8 MB E-Cache) UltraSPARC-III processors were subsequently
added. These nodes are similarly configured and are also connected using Gigabit Ethernet.
In July/2002, two Sun Fire 15K servers were added each with 72 x 900 Mhz (8 MB E-Cache) UltraSPARC-III processors and 144 GB of RAM. Each Sun Fire is equipped with both 10/100Mbps and 1000Mbps (Gigabit) Ethernet Network Interface Cards (NIC's) and multiple
Fibre Channel Arbitrated Loop (FC-AL) host adapters for future Fibre Channel storage connections. Gigabit Ethernet will serve as the cluster interconnect while we evaluate the performance of the cluster and other cluster interconnect technologies.
Interactive logins to the Sun Fire cluster is only supported on the master cluster node (sfnode0.hpcvl.queensu.ca). Regular telnet and file transfer protocol (ftp) sessions have been disabled. You must connect to the master cluster node using the Secure
Shell (SSH) suite of utilities - ssh, scp, and sftp. Our SSH server (sshd) currently support SSH v1, v1.5, and v2 of the Secure Shell protocol.
Click here for additional information on user accounts, security requirements, cluster access, and usage policies.
|
|
Back to top
File Storage: Disk
Disk storage for software applications, user home directories, and temporary space is provided by 11.7TB of Sun StorEdge T3 Fibre Channel disk technology. Currently we have five Sun StorEdge racks with 12x 324GB (9x 36GB FC-AL disk drives) StorEdge
T3 disk arrays. We are now using Sun QFS (v3.5) High Performance HPC system to manage the T3 arrays.
One of the Sun Fire nodes (sfnode0.hpcvl.queensu.ca) acts as the NFS server for the Sun Fire cluster and will serve most of the StorEdge T3 storage. A single StorEdge T3 disk array is connected to each Sun Fire server as local scratch space. This local
scratch space is available to all nodes in the cluster via NFS.
Back to top
File Storage: Archival
We are currently backing up User's home directories for a period of ONE Month after which, the disk space will be recycled.
We have deployed a large DLT (digital linear tape) tape backup system using a Sun Enterprise 450 Workgroup server and a Sun StorEdge L700 (L700) tape library. We are expecting the tape backup system to be operational by August 2001. At this time, we
will implement a backup and retrieval strategy which will include archiving user files located in each users home directory (i.e., all files/directories referenced by the $HOME environment variable).
Back to top
Software Environment
Currently all Sun servers run the 64-bit Solaris 9 Operating Environment. Additional installed software includes:
- Sun Forte Developer 6 Update 2
Sun Forte Developer 6 (Update 2) is Sun's software development environment. Forte Developer 6 includes a complete set of graphical and command line tools to help you build, debug, run, and tune your C, C++, Fortran, and high performance FORTRAN applications.
[more»]
- Sun HPC ClusterTools 4.0
Sun HPC ClusterTools 4.0 is a suite of applications and libraries for high performance software development and workload management of serial and parallel applications. [more»]
- Sun Grid Engine 5.3 Enterprise Edition
Sun Grid Engine software is distributed workload management software that optimizes utilization of software and hardware resources in heterogeneous networked environments. GridEngine is only available on the Sun Fire cluster. [more»]
|
