Is there any reason that Cassandra is the focus of this article? It is really silly and irresponsible to peg a nascent project like that without any reasoning or sources. I'm sure something changed besides just a Cassandra rollout, and wasn't Digg using it on v3 too?
I think Cassandra is pretty well tested. There have been lots of super-large-scale deployments. It just seems lame to blame it on that, but I guess maybe their anonymous sources inside Digg revealed it? But then we'd hope they'd know if the problem was with the datastore or the implementation.
I get that some VP suggested a new buzzwordy technology, they gave him enough rope to hang himself and he did and left the company with a broken pile of crap. It could happen, if you have a healthy company you give trust to people. That it got this far doesn't speak well of the rest of management. It doesn't speak really well of the rest of the team either. Shouldn't there have been some circuit breakers or something?
Digg isn't a poor, bring-your-own-laptop startup. They've got resources, they've had substantial investment. They can afford to build and test software and I know of no real marketing reason they had to push something untested out. Rose could go out and say it wasn't done, it's going to take more time.
How does Rose keep his job? Wasn't he this VP's boss?
And it's single technology? That couldn't have been vetted and tested independently of all of digg? Really? And MongoDB, or hadoop, or one of the dozens of other nosqls wouldn't work, either? You do what you have to do and there is never a 'truth' with VPs and CxOs are canned, but it all doesn't float with me, just looks like another over valued and under talented company, got lucky and the blind squirrel found the nut, there isn't gonna be a second act.
Maybe it just seems low brow to me, name and finger the guy, blame the opensource tool you use, never explain or elaborate why you launched anyways when you were fixing bugs in the tool at the 11th hour.
Perhaps, though I think my sour grapes specifically about digg are somewhat dead. What really makes me angry is the joke that is the VC industry, which is really the root cause of this. In April I talked to an associate at Andreesen-Horowitz (where Kevin Rose is the mayor on foursquare), who said that digg, after many difficulties, was on the right track. From their perspective, engineers are cogs in the machine and the only thing that matters are executives. I worked closely with the executives at digg, and they spent the vast majority of the time feathering their nests, and the digg v4 rollout is just the logical conclusion of them draining value for five years.
Also I said this was a "rough guess." Comments on TC from people I know who worked there (there's been a mass exodus for the last two years) suggest I'm mostly correct.
Finally, I think "sour grapes" is a poor rejoinder. Especially since it's the standard PR line at digg.
No, I apologize. I didn't mean to sound like I was attacking you or suggest that you supported it. It just seemed like a good place to interject. It's such a transparent move on Digg's part.
Part of the reason that Cassandra may be the focus is that Kevin Rose places much of the stability problems on Cassandra's shoulders. Two or three minutes into the most recent diggnation he talks about Cassandra and describes Cassandra as "very beta-stage software" and says how days before the launch at least some of their focus was on fixing "Cassandra problems" rather than issues with Digg v4.
If true, it's a lame excuse: blame should be placed on the people that decided to use the beta software to power their very-important-to-their-paycheck website
So if you watch the Revision3 video where they talk about the stability issues Kevin throws Cassandra out there nothing really specific but when discussing downtime its the only technical factor he mentions.
He calls it "still beta software" and states they were fixing bugs in Cassandra during the days leading up to the release of v4.
In reply to you and epoxy, sounds like a gross failure of engineering management. If you're having problems of this sort of foundational nature a few days before the planned release, it strongly suggests you should delay and figure out what's generally gone wrong with the project.
I think Cassandra is pretty well tested. There have been lots of super-large-scale deployments. It just seems lame to blame it on that, but I guess maybe their anonymous sources inside Digg revealed it? But then we'd hope they'd know if the problem was with the datastore or the implementation.