Eigen is a startup consultancy group (www.eigen.co.uk), with a special focus on security. They took me on (via UNIX Connections Ltd) to assist the MD in an installation of Commamd Post(tm) at East Midland Electric. He trained me up, and steered me through the tightly scheduled staging process.
That meant building a cluster of four HP boxes (series 800 and series 700). Taking them out of the shrink wrap, obtaining floor space, installing the OS, applying the relevent patches, configuring the IP subnet, NIS, NFS, SNMP, disk-mirroring (etc), and installing the platform of Boole+Babbage integration components. Each monitoring some feature of the network, with a demonstration fault and sample display map.
There were occasional hiccups, such as a DCHP server which was (intermittently) allocating our IP address from it's pool -- I added static ARP routes until they fixed it. Plus the usual care needed, whilst wiring up a system to monitor every machine (a dozen HP-9000's and a mainframe), with the computer room being simultaneously re-wired by their internal people.
The MD was busy, so he did the project management / presentations, and I did most of the installation, with the exception of the Sybase kernel configuration (I followed a script), and the configuration of the mainframe, (he did that, then left saying "it's all done, you just have to repeat it for the other IOC box ...")
Command Post is an industry leading Enterprise Management system, which monitors all diagnostic channels over the network, detects errors and tracks them, via a database and various maps. In practice the site has to engineer and maintain their own fault detectors, and subsequent action scripts, as only the local people can take deep control of their network. The phase-one staging (installation) is itself a complicated process, built from many sub-systems which have to be in-place, and reliable.
EME and Perot (a world-wide Facilities Management group, with a strong presence in Nottingham), take their systems very seriously, and have a Systems Management Centre, with a huge wall map (projected X11 display), that shows a logical map of their network (all green for OK, red points for faults) - much like an SNMP map, but for processes as well as network links.
HP/UX 10.20 and HP-9000 can sometimes be a bit fiddly, with lots of serial-keys, patches, and local HP-idioms (HP/UX is not SVR4 based, but often follows similar command switches). HP/UX is a well supported platform, with exciting features like logical drive mirroring, and hot-drive-pluggability (the stomach grinding test is to remove a drive from the running host, smile at the nearest fainting engineer, and put it back). I would be very interested in working with a HP site.
I left them with a fully working system, with backup scripts, delivered to schedule, a site configuration report with several security points to enhance their policy, and a system reporting script that firstly obtained HP/UX system information, then tabulated that information for the 4 boxes into a table which highlighted where system settings had been specially adjusted, in a way that was almost "blindingly obvious".
I also left Eigen with an triple boot W95/NT/Linux box, that neatly fitted into a 1.3G Portable Toshiba, with just enough space to run Remedy, PCMCIA and a DCHP IP configuration. That required a custom written boot-loader, to boot and run unix inside a big DOS file as it's root "partition".
There were other tasks, such as scratch building a Solaris-2.5, installing Remedy, and several days showing the Co-Op bank (Skelmersdale) how to create drill-down maps, with both horizontal and vertical combing of the data, ie select alert by component, select by application, display as <blobs>.