2010/10/01

Back in the Machine Room Again - Logically and Physically Grounded

When I first started out in software as a full-time permanent professional programmer (as opposed to the years of practice as a hobbyist and intern) in 1993, we didn't have much of a server, and Lantastic wasn't exactly much of a network.  But we got the job done and built some great software (in C for handhelds).  We used SourceSafe (before Microsoft bought it) and we had version control and builds and testing and everything which is apparently still not used in some shops these days.

When 1996 rolled around and I entered what you might call the business computing environment, we actually had a "machine room".  It had its own air conditioning.  There were only a couple machines in the room - a huge HP UX machine and a Windows NT 3.51 machine running Tomcat for a BBS/FTP/web server.  There was a rack for telecom equipment, but it wasn't bolted down and the floor was carpeted! (static was always fun)  Over the years, I guided that little room through a new tiling (delft blue was my choice), an expansion (and moving of the door) and eventually to complete retirement as we moved our data center to a dedicated facility (and eventually to another site after that).  I remembered Henry who was working behind a rack one day when a KVM switch fell on his head, leaving a lot of blood on the floor.

For the last three years at the bank, I never got to go into the data center.  Neither data center we used was even in Louisiana.  I didn't miss it that much because I hadn't been in a data center in a while at OCA/Orthosynetics.  We had all become quite detached from the hardware over the years.  I guess that's an unforeseen effect of the cloud.  For other people, I imagine that computing power has been like that for a while.  Those people never had to think about cooling or cabling or power.  I wonder about the kids these days - if the cloud will mean their whole impression of computing and what computers really are will be different.  Computing was never abstract to me.  Sure, there are aspects which are theoretical, but it is always grounded in a physical implementation.

Sometimes people in database design talk about logical and physical designs.  It's so funny, because the physical design is really just another logical design in terms of the vendor database.  And then even that translates to more logical designs, and even storage now is not physical.  And then eventually, after a lot of SAN controllers and stuff, you eventually get to a bit on the disk, which is a logical version of (finally) some magnetic orientation of (physical) atoms.  It cracks me up when people talk about clustered indexes in SQL Server being the order on the disk - dude you are so far removed from knowing what order things are on the disk, you can forget about even talking about disks - you could call them green clovers or blue diamonds for all it matters!

I was thinking about Henry (R.I.P. - I miss you) today when I went into the machine room at my new client engagement yesterday.  First thing I looked at was the A/C - they've got a new unit - very shiny.  It has pressure gauges.  My first question was about the building water supply - open or closed?  I remembered back at Lakeway where we had to ensure that the filters were cleaned regularly of insect carapaces which came into the water in the reservoir at the roof.  Because they have direct pressure gauges, they can see problems like high head pressure directly (we used to get them reported on the Liebert's display).  The racks were not as full as what I was used to and they'll be shrinking with more virtualization.  Their SAN was only part of a rack, and they don't have a switch which takes up its own rack.  Although I'm not ultimately responsible for hardware and network any more, it was great to be able to talk about their network architecture and stay grounded.

1 comment: