|> home > services > storage|
The system has eight I/O nodes, configured across two equipment frames, and whereby each node in one frame mirrors or mimics the equivalent node in the other frame.
There are thus three pairs of VSD (Virtual Shared Disc) servers which control the disc arrays, with alternate pathing such that a VSD server could pick up the actions of its opposite number whilst maintaining its own disc control functions.
Each VSD server contains 16 Power 5 CPUs, 8Gbyte of memory, 4 x FC adapters and 2 x PCI HPS switch adapters. The remaining pair of I/O nodes are configured as the primary login partition and the secondary (fail-over) login partition. The secondary login partition also acts as the primary server for TSM (Tivoli Storage Manager) with the primary login partition acting as the TSM fail-over partition. Additionally, all serial batch jobs run on the otherwise lightly loaded secondary login partition. This sharing of roles enables full redundancy and resiliance to failure for principal service operating functions. The primary and secondary login partitions each contain 16 Power 5 CPUs and 128 Gbyte of memory. The sustained bandwidth between the compute nodes and the discs is of the order of 1GB/s.
External connectivity is provided through the Gigabit Ethernet links.
Archiving functionality is maintained 24x7 and independently of any backup operations which may take place.
The LTO tape drives are attached to the TSM server and the 'standby' server using 6 x 16-port 2Gb switches. Each server is attached to each switch using 2 x 2Gb links. The LTO drives themselves support 1Gb links. This provides a SAN with 2 separate fabrics and some resilience against the loss of a switch due to fabric/firmware problems.
Also see the HPCx Archiving User Guide, V 1.2.
|http://www.hpcx.ac.uk/services/storage/index.html||contact email - firstname.lastname@example.org||© UoE HPCX Ltd|