Hacker News Reader

Comment by dheera | Hacker News Reader

Self hosting being cheaper is obvious. It's just a nightmare to manage reliability. With AWS or any other IaaS provider you're paying to outsource shifts of people around the clock that will work to keep servers and network connections up and running.

Colocation isn't that much better either because there are still situations (like corrupt filesystems or malfunctioning power supplies) that would require physical access to the machine.

With an IaaS provider you're free to go live wherever you want and focus on your business and not need to schedule shifts of people within driving distance of the colocation facility on-call.

Ultimately you're paying to reclaim time, which in almost all cases for startups is a worthwhile trade. Outsource everything that isn't your core business.

View on HN

>With AWS or any other IaaS provider you're paying to outsource shifts of people around the clock that will work to keep servers and network connections up and running.

This is so often repeated, but at my previous company, the amount of internal outages was smaller than the amount of outages on AWS or GCP. We then migrated to AWS for some stupid reason (we didn't need scalability, that's for sure) and we had a problem with redis nodes: the instances they were running on were too small and an upgrade went wrong. We waited until the next shift for the fix. The irony.

That's great, but when you're a startup and your only option is Comcast, outages are a very real thing.

Our first real office also had internet access run by these 2 dudes in the basement and outages every few weeks for a couple hours.

>That's great, but when you're a startup and your only option is Comcast, outages are a very real thing.

Then send your hardware to a colo for a fraction of the cost of AWS

> Colocation isn't that much better either because there are still situations (like corrupt filesystems or malfunctioning power supplies) that would require physical access to the machine.

I disagree.

You still get all those problems with a shared provider, recall Ghandi's outage?, DigitalOcean, Linnode's and AWS a week ago. Your also stuck in a tighter situation when it does occur.

With years of colocation I have never encountered a power-out event within the datacentre. That's a potential seven years of uptime if I hadn't changed providers in-between.

Nor have I ever suffered a corrupt filesystems and that includes myself preforming DR monthly via sudden power-loss cold-boots. Sure I accidentally deleted /etc/ once.

The only issue out-standing is a failed RAM stick however it's a poorly server which I'm retiring in the new year. Poor thing.

iDrac and similar devices can provide “physical” access without physical access.

People have been using them in collocations for years. Collocations also provide hardware swap services. They can replace a power supply for you.

Reply on news.ycombinator.com