Figure 1: The proposed HP UNIX Clusters
specific functions are assigned to particular machines, as indicated in the diagram below. The functions deemed most critical, file-serving and mail, will be implemented first, with additional functions (and their respective machines, etc.) to be added later.
To examine a machine's configuration and proposed usage more closely, click on the machine or its title.
Figure 1: The proposed HP UNIX Clusters
The HP systems will be implemented in phases with some overlap with the existing IBM RISC/System 6000 machines. The implementation is a joint effort between Cal Poly's ITS staff and an implementation team from HP.
HP has proposed a system which provides redundancy in the event of processor failure or overload. This redundancy is generally referred to as "fail-over" (HP MC/Service Guard). Mission critical processes (or jobs) on the failing or overloaded system are started on another system which has redundant access to the first machines RAID arrays. These processes are bundled together into whats known as a package which occupies one or more RAID drive strings of 16 GB each.
This redundancy should limit user perceived down time. While a system going down may have an immediate affect on a group of users, the fail-over site will pick up the load and those users will be able to reinitiate requests which previously failed.
There are basically two clusters of three machines each. The first cluster (the upper machines in the diagram) is to be set up in a high availability group with fail-over following the grey arrows. The second cluster (the lower machines in the diagram) will also be set up in a high availability group with fail-over following the grey arrows. The red lines indicate secondary internal etrhernet redundancy and the blue circle is a 10Base-T Ethernet that is used to by the system to determine that their Service Guard partners are up and operational.
Revised by: George Westlund (gwestlun@calpoly.edu)