Hacker News Reader

Comment by dhd415 | Hacker News Reader

This strikes me as a bit of a straw man -- "AWS has operational skill. Open source companies with managed service offerings of their software probably do not." The operational skills that AWS has are primarily at the infrastructure layer. If my experience with their managed software services and the tons of complaints in their support forums are any indication, their operational excellence drops off dramatically at the software layer. I don't fault AWS for trying to move up the value chain, but bashing on open source companies as "not a viable business model" distracts from the real question of who can delivery a better managed software offering on top of AWS's (or GCP's or Azure's) managed hardware platform.

View on HN

So you’re judging quality by support forum posts? Isn’t that the very definition of a self selected sample? Who goes into a support forum to post about how well something is working?

You can see how the company responds to customer problems though.

Not really, in the case of AWS, if you’re running a business depending on it, you’re not depending on getting help from a public forum, you have at least a business support contract where you get immediate, live support from people who can see your account, do a screen share with you etc. my experience with AWS is that they are batting close to 100.

One of the ideas being kicked around inside the bit of Pivotal that works on Cloud Foundry, back when i was there, was the need for some abstraction that lets you cleanly separate those things. You want your sysadmin/SRE/cloud provider/platform team to be able to do all the hardware/unix/networking stuff, and then people who know the managed software really well to be able to look after them, without having to worry about the rest.

Cloud Foundry (and things like it) does a reasonable job of separating those things for "applications", ie web apps, pretty well, but didn't have an equivalent for data services. Nobody was quite sure what it should look like.

I suppose part of Kubernetes's success has been that it looks like a solution to this. I'm not sure how well it really works, though.

If something like this worked, it would technically enable a marketplace model, where AWS or GCP do their efficient hardware thing, then the software experts provide a value-added service on top of that. What we would then be missing is a way to economically enable that model, where it doesn't make sense for AWS etc to just provide their own version of the service and capture all the profit.

Reply on news.ycombinator.com