Rogue Nodes
Description:
The rogue nodes are 50 IBM Netfinity workstations
with 768M of memory and between 2 and 4 80GB disks.
They are connected with 100mbit ethernet for management
and a research 100mbit channel bonded ethernet connection.
The rogue nodes are accessable only from within UMIACS/CFAR/CS.
Off-site users will need to connect through redleader
to gain access. |
 |
| |
Node |
Aggregate |
| Processors: |
1 x PIII-650Mhz |
50 |
| Memory |
768MB |
38.4GB |
| Disk: |
01-40 2x80GB
41-50 4x80GB |
9.6TB |
| Interconnect: |
100mbit fast ethernet (rogueXX)
200mbit channel bonded ethernet (rogueXX-b) |
|
| OS: |
RedHat 7.3 |
|
Reservations:
There is no scheduler or reservation service on the rogue
nodes. Common practice is to send e-mail to cluster-forum@umiacs.umd.edu
stating that you need sole access to certain resources during
a time frame.
Disk Use:
The internal disks are mounted as /scratch1 - /scratch4.
Please note that these disks are NOT backed
up, and no attempt will be made to retrieve data from failed
disks. The disks are quite prone to failure since they are
the IBM deskstar disks that had a recall done on them. Your
home or group project directory should be used to store
data that you care about, please do not do any massively
I/O intensive work out of your home directory since it is
a shared resource. |
