I have a few friends that were avid users of Ma.gnolia, which was a social bookmarking tool, similar to Delicious.
I had an account there, but never really used the service much. I say similar to Delicious and use the past tense because Ma.gnolia’s database crashed on January 30. They recently announced that all user data were completely irretrievable and lost forever.
The people I know who used the service were resigned to a full fail, and it seemed pretty apparent in the first few days after the crash that getting any data back would be a lucky thing. To his credit, Larry Halff, the founder of Ma.gnolia, was forthright in admitting the mistakes they’d made, most notably with backups. He sat down with Chris Messina to chat about the fail in this vid/podcast.
Not that his “my bad” makes the fail any less annoying to users, but it’s refreshing and honest. I admire this trend in responsibility; Jason Kilar, CEO of Hulu, provides another example. This is fodder for another time though.
The point here isn’t to snark about how Web 2.0 startups aren’t reliable. For the most part, they are, at least in direct proportion with their cost. Before you jump on Ma.gnolia for running their operation on four Mac Minis and two Mac servers, keep in mind their service was free (as in beer).
Economies of scale and technology advances have allowed ideas to become reality much faster than even 10 years ago. Imagine Mag.nolia’s service built in 1998.
You’d need big iron in a datacenter and a bunch of developers. Maybe the service would be more reliable, but at what cost? My guess is that a service like that would have required at least $10 million in venture funding to get off the ground.
I don’t know how much, if any, money Ma.gnolia took in from VCs or advertisers. It’s a general point.
Running a reliable service requires money and lots of it. This is why Facebook has spent several hundred million in VC money on its infrastructure. This is why Google has spent billions on its infrastructure and has more than 30 datacenters. Imagine the revolt if Facebook lost even 10% of its user data (say for about 17 million people), or if Google lost 10% of its index (only about 100 billion web pages).
So, what have we learned? Don’t rely on a service too much because failure happens, even if you’re prepared. Speaking of which, do you back up your data?
Ma.gnolia may have had a faulty backup scheme, but at least they were trying.
So, do you rely too much on free services? Maybe you use Flickr for free, but do you have those pictures backed up somewhere?
How would you feel if your Web 2.0 service of choice lost your data? Find the comments.
An interesting post, I didn't use magnolia but I think it just goes to prove 'on the internet no-one knows you are a one-man band operating from your kitchen'.
Personally, I think Larry Hallf made a classic error – he didn't realise his areas of weaknesses. I think any half-decent MySQL DBA should have been able to assist with a resilient backup and recovery strategy for a database that was only a mere 500GB. I also suspect he could addressed the issue for a lot less than $10 million.
I use a lot of Web 2.0 services, primarily from Google and yes, they have occasional failures – Gmail was down this week for example – and although I get these services for 'free' as in no money, I increasingly worry about the price I am paying entrusting so much of my data to Google.
Of course, even Google with all their power, employees and expensive data centres, are not exempt from reviewing their portfolio of services – I was really disappointed when they recently decide to ditch Notebook although there are multiple export options to alternative services.
Basically, I don't entrust anything to the cloud that I couldn't afford to lose. Fr example, I store photos in Picasa but all my photos are also mastered on my hard drive. When did I last back them up to my external HD – why last October, of course 🙂
Preach the backup gospel.
I agree fully that relying on the 'tubes is asking for trouble. As you say, even if failure doesn't get you, shuttering the service might.
Good thoughts, thanks.
When you mentioned the mere possibility of $10m VC funding, I could have sworn I could hear three Exadata salesmen beating a path to Larry Halff's door !
Heh, I wonder if he jumped from MySQL to Oracle after the fail.
“Heh, I wonder if he jumped from MySQL to Oracle after the fail.”
Not unless he was willing to ditch the Macs too.
Sure, I get it. VMs are an option, albeit not the best one, but still. Didn't realize you were a Mac apologist.
Classic case of being a victim of yr own success.
So even if they had competent DBA brains available from the start and assuming MySQL can cope with the volumes (probably) – explosive rate of growth can still nobble you (esp when yr backup reco strategy is um, um um…)
Yeah, makes you wonder how many other little startups out there that are moderately successful are “at risk”. For example, I'm always a little more confident when I see they use AWS.