Current events
From Vd
Contents |
HPC Cafe?
Anyone interested in a HPC Cafe? The HPC staff at USIT could be interested in showing up in on of the Campus Cafes if someone would want to meet us for a cup of warm brown brew. Suggestions may be sent the hpc-core@usit mailing list ;)
Current known issues on titan
- None; Report it if you encounter one.
Scheduled events
Maintenance stop at Wed May 14th
A bug in the core ethernet switch, the Cisco catalyst 6509, enforces a maintenance stop on TITAN 09:00 - 15:00 Wed May 14th. We strongly regret this.
- The queuing system (SGE) was upgraded. - SGE is the [titan|TITAN] queuing system
- The parallel file system (GPFS - General Parallel File System) was upgraded. - GPFS is a global file system that exist on all TITAN computers, servers and compute nodes. GPFS is used for storing home directories for all users, for storage of large amounts of static data and it is used for common scratch space by projects that has such needs.
- The switch (Cisco 6509) controller card was replaced. The 6509 is the TITAN core switch connecting all nodes and servers together. GPFS runs on top of the 6509. The magnum InfiniBand switch will take much of the workload off of the 6509 when it is fully operational.
- All DELL servers had firmware and Linux kernel upgraded.
- The TITAN large servers (SUN 4600) was configured with faster /work (striped over SATAblades). We have connected 4 [SATAblade]s to the 8 SUN Fire X4600 large memory compute nodes in TITAN to speed up writing of large temporary files and checkpoint data. The X4600 has at least from 32GB to 256GB memory.
- The magnum switch was connected to:
- one rack of compute nodes
- 96 new blade based nodes
- by InfiniBand cables
- NorStore servers was connected to storage devices (ST6540) and to the 10Gbit ethernet (CX4 technology is being applied in NorStore while Tier-1 storage resources use SR/LR).
Upgrade of all computer nodes from dual core to quad core CPUs
- Date not set, see TITAN3 upgrade
Installation of InfiniBand on all computer nodes
- Date not set, see TITAN3 upgrade
TITAN2 -> TITAN3 major upgrade
- Date not set for the differen activities, this will happen before summer 2008
- The newly arrived 96 new compute nodes will constitute TITAN3
- The 96 nodes will be the first to be connected to the new magnum InfiniBand switch
- The 96 nodes will be connected to new GPFS-home directories (TITAN3-home)
- The 96 nodes will be connected to a new GPFS-scratch
- Some projects will be moved from TITAN2 to TITAN3
- 3-4 TITAN2 racks will be upgraded from dual core to quad code and get InfiniBand and will be migrated to TITAN3
- (Step 3 will be repeated 3-4 times until all compute racks are migrated from TITAN2 to TITAN3)
Old events
- December 11th
- Scheduled downtime 08:00 - 16:00
- GPFS upgrade
- ...