Slashdot Mirror
The New IT Crisis
Matey-O writes "You've succeeded in delivering 5 9's, your server farm is a well oiled machine, the helpdesk lines lie dormant. No? Well then how do you get credit for the work you do, when all that's noticed is the downtime? When the IT budget has to be justified, and you're overworked, undermanned, and you have to apply three patches to 100 servers before Close of Business, what has to change in IT before we melt down? Marc Andreessen has an interesting article on what has to happen to IT next."
Ok seriously, does anyone RTFA anymore? How about the comments? This is a clear PR stunt aimed at producing more leads for Marc's new company. And ZDnet, that fine bastion of even-handed IT reporting, has once again saved us all by printing only the relevant facts. Just once I'd liek to believe that one of my old IT heroes didn't sell out and become a corporate whore (can you say RMS anyone?).
-- People who think they know it all, really annoy those of us who do!
I've worked in a few large outfits and in my own small business, and I can testify there is almost zero correlation in a large office between the work that you do and people's perception of what you do.
The people who have the most problems, the ones who have a terrible catastrophe which just always seems to happen to them, are seen as the problem solvers. Despite the fact that their own lack of organization, incompetence, or laziness often brings these things upon them. No matter, they can proudly trumpet how they once again "saved the company" and worked 30 hours straight. The ones like me who prevent the problems, who organize their day so that nothing exciting happens if it can be avoided, and quietly solve problems on their own without assistance before people notice them, are seen as either invisible or lazy.
And no, I don't work a 30 hour day. Ever. I don't need to. I'm not bitter... gak!
The baby's fine -- please stop sending business cards.
I think a better point to critique on his phone analogy is the implied point that the phone system isn't held together with "bailing wire" or "chewing gum." Nope, it's all pretty standardized, well-integrated equipment. Why is the phone service so much more "professional" than IT services?
;^)
Because phone service is a relatively well-defined, consistent, limited problem domain. Internet servers, dynamic web sites, and local security are loosely-defined, constantly shifting, open-ended problem domains. They're very different, and you can't compare one to the other.
However, for certain applications, there are well-defined standards, and well-defined practices. Still, for a lot of IT, it's a matter of custom engineering and architecture. For example, online content management: you can buy one of the management engines off-the-shelf, which will probably do most of what you need in a structured manner. For CRM, well, there's about a dozen of those. These packages are well-behaved in that it provides a well-defined interface, but that's not always an option (i know, i used to do data migration for small- to medium-sized businesses. at the low end, when you change systems, you'd better damned well know perl or some other text processing language to massage the data--that is, you need to be good with your bailing wire).
In the future, this situation will hopefully be better with standardization (mostly using XML it seems, even though the actual encoding doesn't really matter.. we could have standardized years ago, but nobody saw any benefit then). Having done data migration in the past, i'm all for keeping things disparate and non-standard, but that's because the work pays well and is fun
A better analogy might be a pool of corporate autos. Except that you don't have to interconnect the cars to get them to share load dynamically, or access content generated on one to form a report on the other, etc. A lot of IT is like trying to drop a big old hemi into a metro, or getting a suburban to go anywhere with just metros to provide power (two in front and one in back, it might go up a hill!).
Overall, I was not impressed with this article, but I'm afraid it's going to carry more clout than it should. oh well.
I am not typically supportive of
No doubt. Andreesen is trying to fabricate a crisis so he can save his worthless company, Opsware (LoudCloud), from insolvency and his career from the where are they now file.
Someone you trust is one of us.
One of the biggest problems as I see it is that management dont appreciate how important IT has become in their company. Looking at my company, I think they still relate the IT department to the same IT department of five years ago when to be fair the technology was a little easier to grasp and there were much less computers in offices.
Take networking for example - it used to be in our place BNC and the occasional run of UTP cable - all attached to relatively unintelligent devices. Now its all Cisco switches, fiber and cat5e - and it really is a full time job in itself managing a large network with so many 'intelligent' devices.
Also taking into account the addition of so many more servers (SQL, Mail, Finance stuff, Student Records, DNS, Proxy..) - the list is endless. Again, these systems have really bloomed in the past 4 or so years, at least for where I work.
I guess they dont see how much goes on behind doors when it comes to this business..
"Hey! Unless this is a nude love-in, get the hell off my property!!"
Why don't we just go back to the mainframe way of doing things?
When users had dumb terminals on their desk, they had the illision of a full power computer, but it was actually a small box with few moving parts that was linked to high-powered computers (or cluster of computers) who were actually doing the heavy lifting. Since all of the functional components were in the Computer Room, there was rarely a need for tech staff to touch the dumb terminals, and the tech people could work in their own distraction-free environment.
What's more, failures could sometimes be abstracted away from users. Hard drive failures happen, that's a fact of the technology. However, if a HD fails on a user's desk, it means that user has lost the use of their computer until it's fixed. If an HD fails in the datacenter, there's usually a backup of the data which can be put into play immediately by mounting the backup on a good drive that's already spinning. To the users, the disk crash can be practically invisible.
There's already tech technologies such as X-Windows and Windows Terminal Service with which to create GUIs on a dumb terminals. Why does the common secretary need a full-powered PC on her desk anyway?
If you haven't checked into nagios yet, you owe it to yourself to do so. Now. It's a monitoring application that can take action on problems. That's the first step to automating things in the datacenter. It's open-source, and it's highly useful, if a little tricky to get working.
I moved out of a group running a lot of big Sun machines (I set up an E10K for them) because of managerial issues. Before I left, we had a budget item for about $250,000 to set up a monitoring and job-scheduling application. It was going to take *another* Sun box to run, and we were being told that it would take 3 months to get it all set up and configured.
With Nagios, I can do everything we they were talking about implementing. I spent 3 weeks, and it cost me nothing. I employed a dual PII 266 that was collecting dust. (I also used an old P166 as a dedicated kiosk for showing the web page.) My boss and my co-workers think it's great. I'm dying to show it to my old group...
Acts 17:28, "For in Him we live, and move, and have our being."
God.. I pity you peeps..
Seriously, any time I go to work for a new company, the first 2 months are spent automating everything across the board. I'll generally get no sleep quite a few nights writing software, and when I'm done, I just tell management that because of errors in the past, all work orders will be processed through this new interface. I give 'em an internal web addy and then go play Quake/EQ/etc. for the rest of my time there (which has never been less than 1 1/2 years which I get vested options from). The websites handle all the dummy checking, logs all processes to another system to check periodically to make sure all works well, and performs whatever request they want.
I order spares for about every piece of equipment in the building including spare switches and 1-2 spare servers for the occassional *frying* motherboard. If the servers are set up in non-redundant fashion, I make sure load balancers (or happy cisco 6500) are ordered, every server has a backup, all backups are automated and working properly, and put in the DR (Disaster Recovery) report JUST incase they want to go that route.
Frankly, I have MAYBE 1 hour of downtime a year, and that's usually attributed to my tripping over a cable that some numb-nutz (who's gonna get chewed out for an hour) layed outside of the wire maintenance tray. Only reason I move jobs is because I just get WAY too bored. Admining is easy for those who aren't inept.
-What have you contributed lately?
Back in the early 80s (when Marc was in what grade?) the business press ran articles about how software was getting so good that soon we wouldn't need programmers, because writing software would itself be handled by software.
I predict that we'll have software that can write major software at just the same time as we have software that can write convincing novels. In both cases you have the task of putting together language that respresents a broad swath of messy reality.
Now, systems administration may be more like writing a good technical manual than writing a good novel. Ever notice how many good tech handbooks there are out there? You haven't? Maybe it's because novels are easier. Good systems administration is about leveraging people strengths with machine strengths, and vice versa. Automation without the human element is as uncompetitive as, well, the human element without automation these days.
"with their freedom lost all virtue lose" - Milton