Pixar Studio Stories: The Movie Vanishes
youtube.com
youtube.com
I couldn't watch the video because I'm on a train - however, I worked at Pixar from 2002 - 2008, so I'm pretty sure I know exactly what this video is about.
This story was legendary inside the studio - I never thought it would get outside the gates. In some ways, the fact that this had happened in the past, when the assets for a film might actually fit on a single workstation, was great.
Everyone was more careful. The Systems group was great at backups, recovery, and did an amazing job the whole time I was there, and I'm sure this story always sat in the back of their minds as a reminder of why they spent all the extra time and money.
The one time I accidentally locally deleted a day's work, I had a recovery in my home directory in less than an hour.
But the folks at Pixar wanted to make movies. So what they did was convince him that by doing work on movies and commercials, they could advertise the capabilities of their tools.
Eventually, they were able to buy enough time to get their work recognized and the Toy Story deal going which validated their vision and won Steve Jobs over to the movie studio idea.
When it turned out Disney wanted to do a screen release, there was a mad rush as the first iteration of the movie wasn't that great and Pixar had to get John Lasseter off his break to "fix the film"
(there's an interesting story how they basically re storyboarded the film in a weekend with sharpies)
The rest as they say is history. The second film was more successful than the first at the box office.
EDIT - thought I'd share a source. The Pixar Story, you can watch the appropriate part here.
http://www.youtube.com/watch?v=wjrPCIDqbwQ
Starts at 4:30 in
Not fun at all. The kind of pain that brings a grown man to tears.
I've heard, experienced (for a client) and read it so very many times. In fact someone told me that pretty much all backups are useless unless you test them frequently (not just once!).
There are lots of tools that do backups - not so many that check if the backup actually works. This might be a business opportunity for someone.
It's not enough to check if the files wrote successfully - you also have to check that you actually backed up the files you need. And that you didn't miss any.
Something like a full restore with unit tests to make sure you actually have the data you care about.
We tried backup software but each one stored it in a proprietary format and we had no way of checking files to see if they were valid or not. And testing a restore was way too much work. USB drives work because I can see the files, I can open the files and that's good enough for me. KISS
Simple GPG encryption or even truecrypt containers work wonders.
I am vehemently against take-home-backups. You may trust your team members but they can still be robbed, get their cars/homes broken into, or just simply lose the drives. While chances of smart-criminal exploiting your data to gain competitive advantage is very low, the chances of your private information being sold to others is very high - someone can get your customer CC#s, employee SSNs etc. very easily. How often do we hear about an employee lose a laptop or HDD that contains 2 million CC#s and SSNs in an Excel or Access file? Please don't set yourself up for that.
With respect to backup software, my favorite way of backup on Windows is robocopy. I've been using it for 4+ years and it has worked flawlessly EVERY SINGLE time for me. I have very simple DOS scripts that mirror the entire source folder to destination\YYYY-MM folder. Since I only have about 300GB of data and 8TB of storage, I also mirror the source to destination\weekday folder. What this gives me is the ability to retrieve files from any day in last 7 days, or any past month (depending on disk space available). And destination\YYYY-MM always has last night's data.
DOS>robocopy "\\SrcServer\SrcFolder" "\\DestServer\%weekday%" /E /R:1 /W:3 /PURGE /COPYALL /FFT
DOS>robocopy "\\SrcServer\SrcFolder" "\\DestServer\%yyyy%-%mm%" /E /R:1 /W:3 /PURGE /COPYALL /FFT
If you have databases, dump the DBs once a night into one of the SrcFolders. Now you have pretty good backups of your DB too. HDD space is so cheap. I have 3 separate DestServers where I backup to. After the initial backup (and once a month), it's pretty fast each night. If you want remote backups but don't have the bandwidth to copy 20GBs a night, Open-E SANs work great for that. Of course, you can also use DeltaCopy (which is just Rsync for Windows).Our clients are big companies, we don't have nor use credit card machines. Strictly purchase orders and 5-figure checks each month. We are in another country and SSNs aren't used the same way like they are in the US (tied, to your bank, credit, retirement, etc). The most someone could do with an employee ID is buy products with our tax id, we'll happily take a VAT refund for you, or someone could use your SSN to apply for a job (can be problematic around tax time but not end of world).
We do have 2 NASes (one backs up the other) that do dailies. The USB drives are for office on fire/big theft tragedies. I'm trying to get us on cloud storage, but pushing data over a DSL connection was frustrating for just 20 gigs. And the connection died over a weekend. Far easier just to do the USB route. You still can't beat bandwidth of a local HNer living 10 minutes away with a bag of drives.
The gov site visitor was running one of the other programs and "wrote us up" for taking up too much time / effort with backups. we protested and got ourselves cleared to continue. Eight months later I was told that one of the sites lost all their data in a flood. Guess keeping all your data onsite in the basement doesn't really work.
It seems like you could just swap them daily, I guess.
I posted a little about this http://news.ycombinator.com/item?id=1246745
Why would you give away your life's work to some thieving bastard on the internet who you'd never met and didn't care about you, anyway?
On a large operation you can't rule out the bitter ex-employee factor either.
If you have a guy working for you who would leak something you're working on, the solution isn't to not allow people to take things home, it's to fire people who would harm your company.