Nyetwork
« Moving the earth. »
Lately, I begin to look forward to Friday by late Monday afternoon. The weeks pass slowly and the weekends fly by in a flurry of laundry, cleaning house and dog walking. My project from hell at work just keeps grinding on and on and on, wearing down my patience for and my interest in technology. Or, at least, the janitorial side of technology. After thinking about the jobs I've had over the years, every single one has had at least one guy who knows just enough to be dangerous and subsequently breaks systems in new, previously unimagined ways. Clean-up on aisle 6. Bring the mop.
The introduction of the Solaris Service Management Facility (SMF) in Solaris 10 is a wet dream for guys like this as it is both new and powerful enough to lay waste to a fresh system in less than an hour. When I think of the SMF, I have a mental image of a committee of engineers hovering around the unholy birth where there emerged a hairy, middle-aged, very unattractive guy to the horror and delight of all. Never before have I seen a system that asks for the system maintenance password on the console, only to keep right on booting to multi-user with services like cron and syslog somehow in a 'disabled' state. I still have no idea which combination of services this person disabled in a spate of paranoia over security to achieve this level of system fuckitude and I'm not entirely sure that I want to know. Soon we'll see a Solaris 10 system exploit that quietly disables the 'boot' service, thus rendering a system unbootable on the next reboot. Nifty.
I'm sure that the engineers at Sun thought it was a brilliant idea to streamline the somewhat cryptic and mysterious system processes for the pretty GUI jockeys who wouldn't know a command line if it bit them in the arse, but what used to be simple is now a baroque labyrinth of XML, rc scripts with 'methods' and databases that have the power to cripple your system without even trying. The services will list their dependencies, yet won't warn you what services depend on THEM so that when you disable a service you can, in fact, unwittingly disable half of your system. What's the point of having a database with dependencies if it doesn't work in both directions? I suppose that feature is only available in the java-based management GUI that I can't use.
Here it is 2006 and we're still using metal tape, have UFS filesystems that max out at 1 inode per MB in filesystems over 1 terabyte, have OS tools that are essentially worthless on filesystems over 1 terabyte, etc. and they're redesigning something that has worked well for over a decade? I'm sure ZFS is going to be ready for production any year now. Sure. The folks who buy into the whole 'Web 2.0' and internet 'revolution' would probably be a lot less enthusiastic if they saw the bowels of the systems that are still remarkably similar to the big ugly boxes from the 1970s, if only a whole lot smaller. The 'revolution' is how much money people are making off the suckers who have bought into the hype. It's amazing that Google is worth billions while running on a platform that has not improved dramatically over the past two decades, we've just gotten better at making it look more sexy on the user end. A large number of cheap PCs clustered together is still just a bunch of cheap PCs. Too much money has been invested in infrastructure and in software and vendors are going to milk it until there is no other option but to move forward. It's much like gasoline powered automobiles where ubiquity, infrastructure and financial interests have kept us driving cars with the same or less fuel efficiency and the same form factor for the past 60 or more years. It seems only crisis will force a revolution in design.
I've been reading Company lately and have been enjoying its satire of American corporations. At one point in the book, management lays off all of IT and the network goes down.
Throughout the building, Zephyr Holdings is slowly getting back to full operating speed. Not because the network has been fixed; oh no. The east wing of level 19 remains a barren wasteland. No server lives there. No hub can flourish in 19's harsh, inhospitable conditions. Dry, gasping network cables search for data they will never find. IT is dark and dead and will not recover.
But there is work to be done, network or no network. Two weeks ago the network went down; soon after Senior Management assured the company it would have the problem fixed within a few days; now everyone is realizing it is never going to happen. Work-arounds are springing up everywhere you look, like new grass after rain. In the absence of e-mail, employees are discovering the art of speaking into phones. They are realizing that discussions that previously required three days and six e-mails can, with phones, be settled in minutes. Spam and computer viruses, both of which IT claimed were unsolvable problems, have vanished. The plague of e-mail jokes, funny at first and then not, has been eliminated. The pressure to forward chain letters under threat of personal catastrophe has lifted. In-boxes no longer fill with desperate sales pitches from co-workers trying to shift their cars, or kittens.
To transfer documents from one location to another, workers tighten their shoelaces and stretch their legs. People pass each other in the corridors, papers in hand, exchanging happy greetings. Their brains dizzy from unexpected exercise, they stop to chat and laugh. No one realized there were so many people in Zephyr. Until now, you never saw them. Until now, most people arrived at work, planted their buttocks in a chair, and the twain didn't part until five thirty. Now the corridors are like maternity ward waiting rooms, filled with excited voices and good cheer. Lower-back pain is clearing up. Color is rising. Workers find each other more physically attractive. And nobody receives suspicious looks for leaving the department anymore, not so long as they're clutching a sheaf of papers.
Network - what was that thing ever good for?
What, indeed.
pic via
permalink Ω 21 March 2006, Helsinki






