http://www.joyent.com/blog/on-cascading-failures-and-amazons...
But then, opinions change over time.
And I don't think EBS ever had an outage as severe as https://news.ycombinator.com/item?id=2269329
Disclaimer: everything I know about Hoffman and Joyent I read here. I've never done business with them; I'm pretty sure I should be glad about that.
Do you really want to deal with all the hardware yourself though? Juggling failing disks, dealing with system upgrades, managing edge of the network, debugging strange one-off issues that sometimes require firmware updates, backup power systems, etc. ? Any cloud provider is better/cheaper unless you need multiple racks and hire people to manage them.
> "everyone that’s there talking, [...] - none of them has ever run infrastructure. [...] And there’s a few vocal people that do and have done stuff"
What does this even mean? So do they have experience or not? (in reality majority doesn't, but then again majority doesn't set the course for the project during the summits)
> "Until someone contributes the pieces of the stack that OpenStack needs to become a real platform, Hoffman doesn’t see it amounting to more than the Unix-based Common Desktop Environment did as a Windows alternative in the 1990s."
What pieces exactly? "accessibility" and "platform features"? They're not pieces, they don't exist in isolation. You can't add an accessibility piece, just like you cannot take a product and add security. Also, how can a desktop environment be an alternative to an operating system?
> "a lot of the internet-facing and webscale companies"
Did he really say that in a serious context?
> "all of a sudden be driving by AT&T Park and everyone’s tweeting and Instagramming during the game and your car stops working"
If you design your car in a way that the basic functions (engine, breaks, steering, etc.) compete for the same resources your IM app uses, that's a problem with basic design, not with connectivity.
This article really makes me angry. It's a load of bullshit and who-should-do-what that's either completely incorrect or left without any explanation about the reasons.
</rant>, sorry
Yes, I agree. Problem is that cars are turning into pseudo entertainment centers on wheels. You can only put so much tech into a car until the costs of doing so becomes a burden. Then its just more cost cutting measures. Which might include having one on-board computer powering all of the systems.
Uh, Rackspace has built completely functional and running OpenStack infrastructure, multiple times over (many datacenters).
Its flexibility and desired modularity are its weaknesses. OpenStack means something different to everyone you talk to. I believe this is a direct consequence of its governance model, and lack of a benevolent dictator for life. It's decision-by-committee hell. Jason alludes to this:
> “You have a process that’s like governance, governance, governance, governance, governance. Run VMs, run VMs, run VMs. And what the process needs to be about it accessibility, accessibility, accessibility, accessibility, accessibility, accessibility. Platform features, platform features, platform features, platform features.”
I, too, expect the steam to run out. Of course, I thought that the first time I had the misfortune of working with it.
