Hacker News Reader

Comment by sayhello | Hacker News Reader

I've used linode for a short while and am still using slicehost. If your apps are disk I/O heavy, go for slicehost, if you're looking for more RAM, linode may be a good choice. My stuff was not CPU heavy at the time, so I have no comments on that, but linode seemed pretty fast.

I left linode because I kept getting emails from linode support about disk I/O, and that was for an initial import of a 50GB postgres db.

After that, I kept getting emails from them once in a while, when my app was getting disk I/O busy. I can't run a business with these worries in my mind.

Slicehost is and has been working great for me, and they have great support.

View on HN

Is this the email that states:

This is not meant as a warning or a representation that you are misusing your resources. We encourage you to modify the thresholds based on your own individual needs.

You bailed out because of that?

Haha, you beat me to it! :(

No. Its emails from human beings. Perhaps I got bad nodes... but I had just short of 10 slices if my memory serves me right

Linode runs RAID-10 arrays in all their hosts, and their disk I/O performance reflects that (it's very fast).

As davidw pointed out, those e-mails were generated by completely customizable thresholds. If your default usage results in disk I/O rates higher than the default settings, then you should just raise the settings. The warnings are very handy if your node starts getting starved for RAM and begins to swap stuff in an out a lot.

If their I/O really is significantly better than mine (and I hear that it is often enough that it is quite possibly true... I do know that my I/O does kindof suck) then it's more than just RAID-10. I use raid 10 on all my new servers, too. But, I don't email people who use too much disk I/O. I should do that, really, as I/O is /really hard/ to effectively limit automatically. (yes, yes, IOnice. but it doesn't work very well.)

As I pointed out to davidw, I was not talking about generated emails. It was emails from humans, including from caker himself.

I have no idea what kind of storage strategy they are using, but if I got support tickets filed because i'm doing a db import, well... they're doing it wrong.

Reply on news.ycombinator.com