Hacker News new | past | comments | ask | show | jobs | submit login

I've worked for state government. No amount of failure for any amount of money paid would surprise me.



Curious, what's the biggest amount for failure you've seen during your time in state govt?


http://www.statesman.com/news/news/state-regional-govt-polit...

Texas had a contract with IBM to consolidate all of the servers of all state agencies. I believe it was for $863 million dollars and 7 years. The first thing IBM was supposed to was to replace any hardware that needed replacing. They never replaced anything. A ticket to reboot a server took a week. As nearly as I can tell, IBM just underbid the contract, and figured that the court battles would be cheaper than making good on the contract. The on-going legal wrangling was literally constant. We just compiled our lists of grievances throughout our workdays and forwarded them on to our manager, where presumably, they got all aggregated again and eventually made it to the legal teams.

It was pretty horrendous. You could have developers or production at a dead halt for a week literally because a server (managed by IBM) needed to be rebooted, and they wouldn't do it. Managers would call and rage. I worked at a large agency, and we eventually got administrator access to the machines, which I think we were supposed to use only for read-only access... occasionally, we'd get told to actually reboot them or something. My impression was that IBM would be logging such infractions and just try to use them as more evidence of Texas violating the contract or something. At the very least, we weren't really supposed to use the access, but we did.

I wondered how the whole agency functioned with the IT infrastructure that badly planned. If you needed a new virtual machine, you'd be doing pretty well to get your request in 6 months in advance, literally. For them to click through creating when you literally know they have the capacity (because we had some insight into the infrastructure.) One of my managers went to some session, and there were more than 35 process steps to getting something like that done, if I recall corretly. Any step stalls, that's a black hole; you'd better be calling and checking on requests like that regularly, or they just dead-end somewhere. (Actually, nothing you didn't call about regularly ever got done, including reboots.) I think they got that down to 15ish steps eventually.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: