The SAN layout has been completely changed. There are four ports on each storage controller and four numbers of controllers. There are six numbers of SAN switches in cluster Group-A. The layout of the controller ports and their conenction to SAN switches are shown in figure A.
Table-1 shows the connection of front end servers (ns1 to ns8).
The workstations were accessing the GPFS file system through Ethernet earlier. There were performance issues while doing graphics apps which queried the storage data through ethernet. It was certain that adding a fibre connectivity to the work station which talks directly to the storage via SAN swich would improve performance to a very high degree.
once the w/s and storage controllers were connected via SAN switch, the workstation HBA's were visible in the Storage Manager interface. Then they were added to each of the four controllers.
in the workstation, RDAC drivers were installed and the LUNs sda,sdb,...sdm were visible in the workstation. GPFS daemon once started, idetified these LUns to be part of file system.
copying a test file showed a throughput improvement of 5 times.
some more insights:-
1.The front end and node servers have two ethernet cards, one with 192 Ip and other with 22
IP.
2. All tape drives are in one SAN switch.
3. There are no ISL between the six SAN switches. each of them is independent.Managed by cisco works using a IP assigned to each switch.
4. The fibre HBA has three LED's indicating the speed. Green, yellow and red for 1,2 & 4 Gb/s.
5. If the Tx-Rx mate is correct while plugging in the LC cable to HBA, only one among the three LED glows. If the cable is improperly connected, the three LEd's blink.
